End of training

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+images_0.png filter=lfs diff=lfs merge=lfs -text
+images_1.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,45 @@
+---
+base_model: stable-diffusion-v1-5/stable-diffusion-v1-5
+library_name: diffusers
+license: creativeml-openrail-m
+inference: true
+tags:
+- stable-diffusion
+- stable-diffusion-diffusers
+- text-to-image
+- diffusers
+- controlnet
+- diffusers-training
+---
+
+<!-- This model card has been generated automatically according to the information the training script had access to. You
+should probably proofread and complete it, then remove this comment. -->
+
+
+# controlnet-brity/controlnet-2210
+
+These are controlnet weights trained on stable-diffusion-v1-5/stable-diffusion-v1-5 with new type of conditioning.
+You can find some example images below.
+
+prompt: red circle with blue background
+![images_0)](./images_0.png)
+prompt: cyan circle with brown floral background
+![images_1)](./images_1.png)
+
+
+
+## Intended uses & limitations
+
+#### How to use
+
+```python
+# TODO: add an example code snippet for running this diffusion pipeline
+```
+
+#### Limitations and bias
+
+[TODO: provide examples of latent issues and potential remediations]
+
+## Training details
+
+[TODO: describe the data used to train the model]

README_flux.md

@@ -0,0 +1,430 @@
+# ControlNet training example for FLUX
+
+The `train_controlnet_flux.py` script shows how to implement the ControlNet training procedure and adapt it for [FLUX](https://github.com/black-forest-labs/flux).
+
+Training script provided by LibAI, which is an institution dedicated to the progress and achievement of artificial general intelligence. LibAI is the developer of [cutout.pro](https://www.cutout.pro/) and [promeai.pro](https://www.promeai.pro/).
+> [!NOTE]
+> **Memory consumption**
+>
+> Flux can be quite expensive to run on consumer hardware devices and as a result, ControlNet training of it comes with higher memory requirements than usual. 
+
+> **Gated access**
+>
+> As the model is gated, before using it with diffusers you first need to go to the [FLUX.1 [dev] Hugging Face page](https://huggingface.co/black-forest-labs/FLUX.1-dev), fill in the form and accept the gate. Once you are in, you need to log in so that your system knows you’ve accepted the gate. Use the command below to log in: `huggingface-cli login`
+
+
+## Running locally with PyTorch
+
+### Installing the dependencies
+
+Before running the scripts, make sure to install the library's training dependencies:
+
+**Important**
+
+To make sure you can successfully run the latest versions of the example scripts, we highly recommend **installing from source** and keeping the install up to date as we update the example scripts frequently and install some example-specific requirements. To do this, execute the following steps in a new virtual environment:
+
+```bash
+git clone https://github.com/huggingface/diffusers
+cd diffusers
+pip install -e .
+```
+
+Then cd in the `examples/controlnet` folder and run
+```bash
+pip install -r requirements_flux.txt
+```
+
+And initialize an [🤗Accelerate](https://github.com/huggingface/accelerate/) environment with:
+
+```bash
+accelerate config
+```
+
+Or for a default accelerate configuration without answering questions about your environment
+
+```bash
+accelerate config default
+```
+
+Or if your environment doesn't support an interactive shell (e.g., a notebook)
+
+```python
+from accelerate.utils import write_basic_config
+write_basic_config()
+```
+
+When running `accelerate config`, if we specify torch compile mode to True there can be dramatic speedups.
+
+## Custom Datasets
+
+We support dataset formats:
+The original dataset is hosted in the [ControlNet repo](https://huggingface.co/lllyasviel/ControlNet/blob/main/training/fill50k.zip). We re-uploaded it to be compatible with `datasets` [here](https://huggingface.co/datasets/fusing/fill50k). Note that `datasets` handles dataloading within the training script. To use our example, add `--dataset_name=fusing/fill50k \` to the script and remove line `--jsonl_for_train` mentioned below.
+
+
+We also support importing data from jsonl(xxx.jsonl),using `--jsonl_for_train` to enable it, here is a brief example of jsonl files:
+```sh
+{"image": "xxx", "text": "xxx", "conditioning_image": "xxx"}
+{"image": "xxx", "text": "xxx", "conditioning_image": "xxx"}
+```
+
+## Training
+
+Our training examples use two test conditioning images. They can be downloaded by running
+
+```sh
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_2.png
+```
+
+Then run `huggingface-cli login` to log into your Hugging Face account. This is needed to be able to push the trained ControlNet parameters to Hugging Face Hub.
+
+we can define the num_layers, num_single_layers, which determines the size of the control(default values are num_layers=4, num_single_layers=10)
+
+
+```bash
+accelerate launch train_controlnet_flux.py \
+    --pretrained_model_name_or_path="black-forest-labs/FLUX.1-dev" \
+    --dataset_name=fusing/fill50k \
+    --conditioning_image_column=conditioning_image \
+    --image_column=image \
+    --caption_column=text \
+    --output_dir="path to save model" \
+    --mixed_precision="bf16" \
+    --resolution=512 \
+    --learning_rate=1e-5 \
+    --max_train_steps=15000 \
+    --validation_steps=100 \
+    --checkpointing_steps=200 \
+    --validation_image "./conditioning_image_1.png" "./conditioning_image_2.png" \
+    --validation_prompt "red circle with blue background" "cyan circle with brown floral background" \
+    --train_batch_size=1 \
+    --gradient_accumulation_steps=4 \
+    --report_to="wandb" \
+    --num_double_layers=4 \
+    --num_single_layers=0 \
+    --seed=42 \
+    --push_to_hub \
+```
+
+To better track our training experiments, we're using the following flags in the command above:
+
+* `report_to="wandb` will ensure the training runs are tracked on Weights and Biases.
+* `validation_image`, `validation_prompt`, and `validation_steps` to allow the script to do a few validation inference runs. This allows us to qualitatively check if the training is progressing as expected.
+
+Our experiments were conducted on a single 80GB A100 GPU.
+
+### Inference
+
+Once training is done, we can perform inference like so:
+
+```python
+import torch
+from diffusers.utils import load_image
+from diffusers.pipelines.flux.pipeline_flux_controlnet import FluxControlNetPipeline
+from diffusers.models.controlnet_flux import FluxControlNetModel
+
+base_model = 'black-forest-labs/FLUX.1-dev'
+controlnet_model = 'promeai/FLUX.1-controlnet-lineart-promeai'
+controlnet = FluxControlNetModel.from_pretrained(controlnet_model, torch_dtype=torch.bfloat16)
+pipe = FluxControlNetPipeline.from_pretrained(
+    base_model, 
+    controlnet=controlnet, 
+    torch_dtype=torch.bfloat16
+)
+# enable memory optimizations   
+pipe.enable_model_cpu_offload()
+
+control_image = load_image("https://huggingface.co/promeai/FLUX.1-controlnet-lineart-promeai/resolve/main/images/example-control.jpg")resize((1024, 1024))
+prompt = "cute anime girl with massive fluffy fennec ears and a big fluffy tail blonde messy long hair blue eyes wearing a maid outfit with a long black gold leaf pattern dress and a white apron mouth open holding a fancy black forest cake with candles on top in the kitchen of an old dark Victorian mansion lit by candlelight with a bright window to the foggy forest and very expensive stuff everywhere"
+
+image = pipe(
+    prompt, 
+    control_image=control_image,
+    controlnet_conditioning_scale=0.6,
+    num_inference_steps=28, 
+    guidance_scale=3.5,
+).images[0]
+image.save("./output.png")
+```
+
+## Apply Deepspeed Zero3 
+
+This is an experimental process, I am not sure if it is suitable for everyone, we used this process to successfully train 512 resolution on A100(40g) * 8.
+Please modify some of the code in the script.
+### 1.Customize zero3 settings
+
+Copy the **accelerate_config_zero3.yaml**,modify `num_processes` according to the number of gpus you want to use:
+
+```bash
+compute_environment: LOCAL_MACHINE
+debug: false
+deepspeed_config:
+  gradient_accumulation_steps: 8
+  offload_optimizer_device: cpu
+  offload_param_device: cpu
+  zero3_init_flag: true
+  zero3_save_16bit_model: true
+  zero_stage: 3
+distributed_type: DEEPSPEED
+downcast_bf16: 'no'
+enable_cpu_affinity: false
+machine_rank: 0
+main_training_function: main
+mixed_precision: bf16
+num_machines: 1
+num_processes: 8
+rdzv_backend: static
+same_network: true
+tpu_env: []
+tpu_use_cluster: false
+tpu_use_sudo: false
+use_cpu: false
+```
+
+### 2.Precompute all inputs (latent, embeddings)
+
+In the train_controlnet_flux.py, We need to pre-calculate all parameters and put them into batches.So we first need to rewrite the `compute_embeddings` function. 
+
+```python
+def compute_embeddings(batch, proportion_empty_prompts, vae, flux_controlnet_pipeline, weight_dtype, is_train=True):
+    
+    ### compute text embeddings
+    prompt_batch = batch[args.caption_column]
+    captions = []
+    for caption in prompt_batch:
+        if random.random() < proportion_empty_prompts:
+            captions.append("")
+        elif isinstance(caption, str):
+            captions.append(caption)
+        elif isinstance(caption, (list, np.ndarray)):
+            # take a random caption if there are multiple
+            captions.append(random.choice(caption) if is_train else caption[0])
+    prompt_batch = captions
+    prompt_embeds, pooled_prompt_embeds, text_ids = flux_controlnet_pipeline.encode_prompt(
+        prompt_batch, prompt_2=prompt_batch
+    )
+    prompt_embeds = prompt_embeds.to(dtype=weight_dtype)
+    pooled_prompt_embeds = pooled_prompt_embeds.to(dtype=weight_dtype)
+    text_ids = text_ids.to(dtype=weight_dtype)
+
+    # text_ids [512,3] to [bs,512,3]
+    text_ids = text_ids.unsqueeze(0).expand(prompt_embeds.shape[0], -1, -1)
+
+    ### compute latents
+    def _pack_latents(latents, batch_size, num_channels_latents, height, width):
+        latents = latents.view(batch_size, num_channels_latents, height // 2, 2, width // 2, 2)
+        latents = latents.permute(0, 2, 4, 1, 3, 5)
+        latents = latents.reshape(batch_size, (height // 2) * (width // 2), num_channels_latents * 4)
+        return latents
+
+    # vae encode
+    pixel_values = batch["pixel_values"]
+    pixel_values = torch.stack([image for image in pixel_values]).to(dtype=weight_dtype).to(vae.device)
+    pixel_latents_tmp = vae.encode(pixel_values).latent_dist.sample()
+    pixel_latents_tmp = (pixel_latents_tmp - vae.config.shift_factor) * vae.config.scaling_factor
+    pixel_latents = _pack_latents(
+        pixel_latents_tmp,
+        pixel_values.shape[0],
+        pixel_latents_tmp.shape[1],
+        pixel_latents_tmp.shape[2],
+        pixel_latents_tmp.shape[3],
+    ) 
+
+    control_values = batch["conditioning_pixel_values"]
+    control_values = torch.stack([image for image in control_values]).to(dtype=weight_dtype).to(vae.device)
+    control_latents = vae.encode(control_values).latent_dist.sample()
+    control_latents = (control_latents - vae.config.shift_factor) * vae.config.scaling_factor
+    control_latents = _pack_latents(
+        control_latents,
+        control_values.shape[0],
+        control_latents.shape[1],
+        control_latents.shape[2],
+        control_latents.shape[3],
+    )
+
+    # copied from pipeline_flux_controlnet
+    def _prepare_latent_image_ids(batch_size, height, width, device, dtype):
+        latent_image_ids = torch.zeros(height // 2, width // 2, 3)
+        latent_image_ids[..., 1] = latent_image_ids[..., 1] + torch.arange(height // 2)[:, None]
+        latent_image_ids[..., 2] = latent_image_ids[..., 2] + torch.arange(width // 2)[None, :]
+
+        latent_image_id_height, latent_image_id_width, latent_image_id_channels = latent_image_ids.shape
+
+        latent_image_ids = latent_image_ids[None, :].repeat(batch_size, 1, 1, 1)
+        latent_image_ids = latent_image_ids.reshape(
+            batch_size, latent_image_id_height * latent_image_id_width, latent_image_id_channels
+        )
+
+        return latent_image_ids.to(device=device, dtype=dtype)
+    latent_image_ids = _prepare_latent_image_ids(
+        batch_size=pixel_latents_tmp.shape[0],
+        height=pixel_latents_tmp.shape[2],
+        width=pixel_latents_tmp.shape[3],
+        device=pixel_values.device,
+        dtype=pixel_values.dtype,
+    )
+
+    # unet_added_cond_kwargs = {"pooled_prompt_embeds": pooled_prompt_embeds, "text_ids": text_ids}
+    return {"prompt_embeds": prompt_embeds, "pooled_prompt_embeds": pooled_prompt_embeds, "text_ids": text_ids, "pixel_latents": pixel_latents, "control_latents": control_latents, "latent_image_ids": latent_image_ids}
+```
+
+Because we need images to pass through vae, we need to preprocess the images in the dataset first. At the same time, vae requires more gpu memory, so you may need to modify the `batch_size` below
+```diff
++train_dataset = prepare_train_dataset(train_dataset, accelerator)
+with accelerator.main_process_first():
+    from datasets.fingerprint import Hasher
+
+    # fingerprint used by the cache for the other processes to load the result
+    # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
+    new_fingerprint = Hasher.hash(args)
+    train_dataset = train_dataset.map(
+-        compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint, batch_size=100
++        compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint, batch_size=10
+    )
+
+del text_encoders, tokenizers
+gc.collect()
+torch.cuda.empty_cache()
+
+# Then get the training dataset ready to be passed to the dataloader.
+-train_dataset = prepare_train_dataset(train_dataset, accelerator)
+```
+### 3.Redefine the behavior of getting batchsize
+
+Now that we have all the preprocessing done, we need to modify the `collate_fn` function.
+
+```python
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    pixel_latents = torch.stack([torch.tensor(example["pixel_latents"]) for example in examples])
+    pixel_latents = pixel_latents.to(memory_format=torch.contiguous_format).float()
+
+    control_latents = torch.stack([torch.tensor(example["control_latents"]) for example in examples])
+    control_latents = control_latents.to(memory_format=torch.contiguous_format).float()
+    
+    latent_image_ids= torch.stack([torch.tensor(example["latent_image_ids"]) for example in examples])
+    
+    prompt_ids = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
+
+    pooled_prompt_embeds = torch.stack([torch.tensor(example["pooled_prompt_embeds"]) for example in examples])
+    text_ids = torch.stack([torch.tensor(example["text_ids"]) for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "pixel_latents": pixel_latents,
+        "control_latents": control_latents,
+        "latent_image_ids": latent_image_ids,
+        "prompt_ids": prompt_ids,
+        "unet_added_conditions": {"pooled_prompt_embeds": pooled_prompt_embeds, "time_ids": text_ids},
+    }
+```
+Finally, we just need to modify the way of obtaining various parameters during training.
+```python
+for epoch in range(first_epoch, args.num_train_epochs):
+    for step, batch in enumerate(train_dataloader):
+        with accelerator.accumulate(flux_controlnet):
+            # Convert images to latent space
+            pixel_latents = batch["pixel_latents"].to(dtype=weight_dtype)
+            control_image = batch["control_latents"].to(dtype=weight_dtype)
+            latent_image_ids = batch["latent_image_ids"].to(dtype=weight_dtype)
+
+            # Sample noise that we'll add to the latents
+            noise = torch.randn_like(pixel_latents).to(accelerator.device).to(dtype=weight_dtype)
+            bsz = pixel_latents.shape[0]
+
+            # Sample a random timestep for each image
+            t = torch.sigmoid(torch.randn((bsz,), device=accelerator.device, dtype=weight_dtype))
+
+            # apply flow matching
+            noisy_latents = (
+                1 - t.unsqueeze(1).unsqueeze(2).repeat(1, pixel_latents.shape[1], pixel_latents.shape[2])
+            ) * pixel_latents + t.unsqueeze(1).unsqueeze(2).repeat(
+                1, pixel_latents.shape[1], pixel_latents.shape[2]
+            ) * noise
+
+            guidance_vec = torch.full(
+                (noisy_latents.shape[0],), 3.5, device=noisy_latents.device, dtype=weight_dtype
+            )
+
+            controlnet_block_samples, controlnet_single_block_samples = flux_controlnet(
+                hidden_states=noisy_latents,
+                controlnet_cond=control_image,
+                timestep=t,
+                guidance=guidance_vec,
+                pooled_projections=batch["unet_added_conditions"]["pooled_prompt_embeds"].to(dtype=weight_dtype),
+                encoder_hidden_states=batch["prompt_ids"].to(dtype=weight_dtype),
+                txt_ids=batch["unet_added_conditions"]["time_ids"][0].to(dtype=weight_dtype),
+                img_ids=latent_image_ids[0],
+                return_dict=False,
+            )
+
+            noise_pred = flux_transformer(
+                hidden_states=noisy_latents,
+                timestep=t,
+                guidance=guidance_vec,
+                pooled_projections=batch["unet_added_conditions"]["pooled_prompt_embeds"].to(dtype=weight_dtype),
+                encoder_hidden_states=batch["prompt_ids"].to(dtype=weight_dtype),
+                controlnet_block_samples=[sample.to(dtype=weight_dtype) for sample in controlnet_block_samples]
+                if controlnet_block_samples is not None
+                else None,
+                controlnet_single_block_samples=[
+                    sample.to(dtype=weight_dtype) for sample in controlnet_single_block_samples
+                ]
+                if controlnet_single_block_samples is not None
+                else None,
+                txt_ids=batch["unet_added_conditions"]["time_ids"][0].to(dtype=weight_dtype),
+                img_ids=latent_image_ids[0],
+                return_dict=False,
+            )[0]
+```
+Congratulations! You have completed all the required code modifications required for deepspeedzero3.
+
+### 4.Training with deepspeedzero3
+
+Start!!!
+
+```bash
+export pretrained_model_name_or_path='flux-dev-model-path'
+export MODEL_TYPE='train_model_type'
+export TRAIN_JSON_FILE="your_json_file"
+export CONTROL_TYPE='control_preprocessor_type'
+export CAPTION_COLUMN='caption_column'
+
+export CACHE_DIR="/data/train_csr/.cache/huggingface/"
+export OUTPUT_DIR='/data/train_csr/FLUX/MODEL_OUT/'$MODEL_TYPE
+# The first step is to use Python to precompute all caches.Replace the first line below with this line. (I am not sure why using acclerate would cause problems.)
+
+CUDA_VISIBLE_DEVICES=0 python3 train_controlnet_flux.py \
+
+# The second step is to use the above accelerate config to train
+accelerate  launch  --config_file "./accelerate_config_zero3.yaml" train_controlnet_flux.py \
+    --pretrained_model_name_or_path=$pretrained_model_name_or_path \
+    --jsonl_for_train=$TRAIN_JSON_FILE \
+    --conditioning_image_column=$CONTROL_TYPE \
+    --image_column=image \
+    --caption_column=$CAPTION_COLUMN\
+    --cache_dir=$CACHE_DIR \
+    --tracker_project_name=$MODEL_TYPE \
+    --output_dir=$OUTPUT_DIR \
+    --max_train_steps=500000 \
+    --mixed_precision bf16 \
+    --checkpointing_steps=1000 \
+    --gradient_accumulation_steps=8 \
+    --resolution=512 \
+    --train_batch_size=1 \
+    --learning_rate=1e-5 \
+    --num_double_layers=4 \
+    --num_single_layers=0 \
+    --gradient_checkpointing \
+    --resume_from_checkpoint="latest" \
+    # --use_adafactor \ dont use
+    # --validation_steps=3 \ not support 
+    # --validation_image $VALIDATION_IMAGE \ not support 
+    # --validation_prompt "xxx" \ not support 
+```

README_sd3.md

@@ -0,0 +1,175 @@
+# ControlNet training example for Stable Diffusion 3/3.5 (SD3/3.5)
+
+The `train_controlnet_sd3.py` script shows how to implement the ControlNet training procedure and adapt it for [Stable Diffusion 3](https://arxiv.org/abs/2403.03206) and [Stable Diffusion 3.5](https://stability.ai/news/introducing-stable-diffusion-3-5).
+
+## Running locally with PyTorch
+
+### Installing the dependencies
+
+Before running the scripts, make sure to install the library's training dependencies:
+
+**Important**
+
+To make sure you can successfully run the latest versions of the example scripts, we highly recommend **installing from source** and keeping the install up to date as we update the example scripts frequently and install some example-specific requirements. To do this, execute the following steps in a new virtual environment:
+
+```bash
+git clone https://github.com/huggingface/diffusers
+cd diffusers
+pip install -e .
+```
+
+Then cd in the `examples/controlnet` folder and run
+```bash
+pip install -r requirements_sd3.txt
+```
+
+And initialize an [🤗Accelerate](https://github.com/huggingface/accelerate/) environment with:
+
+```bash
+accelerate config
+```
+
+Or for a default accelerate configuration without answering questions about your environment
+
+```bash
+accelerate config default
+```
+
+Or if your environment doesn't support an interactive shell (e.g., a notebook)
+
+```python
+from accelerate.utils import write_basic_config
+write_basic_config()
+```
+
+When running `accelerate config`, if we specify torch compile mode to True there can be dramatic speedups.
+
+## Circle filling dataset
+
+The original dataset is hosted in the [ControlNet repo](https://huggingface.co/lllyasviel/ControlNet/blob/main/training/fill50k.zip). We re-uploaded it to be compatible with `datasets` [here](https://huggingface.co/datasets/fusing/fill50k). Note that `datasets` handles dataloading within the training script.
+Please download the dataset and unzip it in the directory `fill50k` in the `examples/controlnet` folder.
+
+## Training
+
+First download the SD3 model from [Hugging Face Hub](https://huggingface.co/stabilityai/stable-diffusion-3-medium-diffusers) or the SD3.5 model from [Hugging Face Hub](https://huggingface.co/stabilityai/stable-diffusion-3.5-medium). We will use it as a base model for the ControlNet training.
+> [!NOTE]
+> As the model is gated, before using it with diffusers you first need to go to the [Stable Diffusion 3 Medium Hugging Face page](https://huggingface.co/stabilityai/stable-diffusion-3-medium-diffusers) or [Stable Diffusion 3.5 Large Hugging Face page](https://huggingface.co/stabilityai/stable-diffusion-3.5-medium), fill in the form and accept the gate. Once you are in, you need to log in so that your system knows you’ve accepted the gate. Use the command below to log in:
+
+```bash
+huggingface-cli login
+```
+
+This will also allow us to push the trained model parameters to the Hugging Face Hub platform.
+
+
+Our training examples use two test conditioning images. They can be downloaded by running
+
+```sh
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png
+
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_2.png
+```
+
+Then run the following commands to train a ControlNet model.
+
+```bash
+export MODEL_DIR="stabilityai/stable-diffusion-3-medium-diffusers"
+export OUTPUT_DIR="sd3-controlnet-out"
+
+accelerate launch train_controlnet_sd3.py \
+    --pretrained_model_name_or_path=$MODEL_DIR \
+    --output_dir=$OUTPUT_DIR \
+    --train_data_dir="fill50k" \
+    --resolution=1024 \
+    --learning_rate=1e-5 \
+    --max_train_steps=15000 \
+    --validation_image "./conditioning_image_1.png" "./conditioning_image_2.png" \
+    --validation_prompt "red circle with blue background" "cyan circle with brown floral background" \
+    --validation_steps=100 \
+    --train_batch_size=1 \
+    --gradient_accumulation_steps=4
+```
+
+To train a ControlNet model for Stable Diffusion 3.5, replace the `MODEL_DIR` with `stabilityai/stable-diffusion-3.5-medium`.
+
+To better track our training experiments, we're using flags `validation_image`, `validation_prompt`, and `validation_steps` to allow the script to do a few validation inference runs. This allows us to qualitatively check if the training is progressing as expected.
+
+Our experiments were conducted on a single 40GB A100 GPU.
+
+### Inference
+
+Once training is done, we can perform inference like so:
+
+```python
+from diffusers import StableDiffusion3ControlNetPipeline, SD3ControlNetModel
+from diffusers.utils import load_image
+import torch
+
+base_model_path = "stabilityai/stable-diffusion-3-medium-diffusers"
+controlnet_path = "DavyMorgan/sd3-controlnet-out"
+
+controlnet = SD3ControlNetModel.from_pretrained(controlnet_path, torch_dtype=torch.float16)
+pipe = StableDiffusion3ControlNetPipeline.from_pretrained(
+    base_model_path, controlnet=controlnet
+)
+pipe.to("cuda", torch.float16)
+
+
+control_image = load_image("./conditioning_image_1.png").resize((1024, 1024))
+prompt = "pale golden rod circle with old lace background"
+
+# generate image
+generator = torch.manual_seed(0)
+image = pipe(
+    prompt, num_inference_steps=20, generator=generator, control_image=control_image
+).images[0]
+image.save("./output.png")
+```
+
+Similarly, for SD3.5, replace the `base_model_path` with `stabilityai/stable-diffusion-3.5-medium` and controlnet_path `DavyMorgan/sd35-controlnet-out'.
+
+## Notes
+
+### GPU usage
+
+SD3 is a large model and requires a lot of GPU memory. 
+We recommend using one GPU with at least 80GB of memory.
+Make sure to use the right GPU when configuring the [accelerator](https://huggingface.co/docs/transformers/en/accelerate).
+
+
+## Example results
+
+### SD3
+
+#### After 500 steps with batch size 8
+
+| |  |
+|-------------------|:-------------------------:|
+|| pale golden rod circle with old lace background |
+ ![conditioning image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png) | ![pale golden rod circle with old lace background](https://huggingface.co/datasets/DavyMorgan/sd3-controlnet-results/resolve/main/step-500.png) |
+
+
+#### After 6500 steps with batch size 8:
+
+| |  |
+|-------------------|:-------------------------:|
+|| pale golden rod circle with old lace background |
+ ![conditioning image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png) | ![pale golden rod circle with old lace background](https://huggingface.co/datasets/DavyMorgan/sd3-controlnet-results/resolve/main/step-6500.png) |
+
+### SD3.5
+
+#### After 500 steps with batch size 8
+
+| |                                                                                                                                                     |
+|-------------------|:---------------------------------------------------------------------------------------------------------------------------------------------------:|
+||                                                   pale golden rod circle with old lace background                                                   |
+ ![conditioning image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png) | ![pale golden rod circle with old lace background](https://huggingface.co/datasets/DavyMorgan/sd3-controlnet-results/resolve/main/step-500-3.5.png) |
+
+
+#### After 3000 steps with batch size 8:
+
+| |                                                                                                                                                      |
+|-------------------|:----------------------------------------------------------------------------------------------------------------------------------------------------:|
+||                                                   pale golden rod circle with old lace background                                                    |
+ ![conditioning image](https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png) | ![pale golden rod circle with old lace background](https://huggingface.co/datasets/DavyMorgan/sd3-controlnet-results/resolve/main/step-3000-3.5.png) |
+

README_sdxl.md

@@ -0,0 +1,141 @@
+# ControlNet training example for Stable Diffusion XL (SDXL)
+
+The `train_controlnet_sdxl.py` script shows how to implement the ControlNet training procedure and adapt it for [Stable Diffusion XL](https://huggingface.co/papers/2307.01952).
+
+## Running locally with PyTorch
+
+### Installing the dependencies
+
+Before running the scripts, make sure to install the library's training dependencies:
+
+**Important**
+
+To make sure you can successfully run the latest versions of the example scripts, we highly recommend **installing from source** and keeping the install up to date as we update the example scripts frequently and install some example-specific requirements. To do this, execute the following steps in a new virtual environment:
+
+```bash
+git clone https://github.com/huggingface/diffusers
+cd diffusers
+pip install -e .
+```
+
+Then cd in the `examples/controlnet` folder and run
+```bash
+pip install -r requirements_sdxl.txt
+```
+
+And initialize an [🤗Accelerate](https://github.com/huggingface/accelerate/) environment with:
+
+```bash
+accelerate config
+```
+
+Or for a default accelerate configuration without answering questions about your environment
+
+```bash
+accelerate config default
+```
+
+Or if your environment doesn't support an interactive shell (e.g., a notebook)
+
+```python
+from accelerate.utils import write_basic_config
+write_basic_config()
+```
+
+When running `accelerate config`, if we specify torch compile mode to True there can be dramatic speedups.
+
+## Circle filling dataset
+
+The original dataset is hosted in the [ControlNet repo](https://huggingface.co/lllyasviel/ControlNet/blob/main/training/fill50k.zip). We re-uploaded it to be compatible with `datasets` [here](https://huggingface.co/datasets/fusing/fill50k). Note that `datasets` handles dataloading within the training script.
+
+## Training
+
+Our training examples use two test conditioning images. They can be downloaded by running
+
+```sh
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_1.png
+
+wget https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/diffusers/controlnet_training/conditioning_image_2.png
+```
+
+Then run `huggingface-cli login` to log into your Hugging Face account. This is needed to be able to push the trained ControlNet parameters to Hugging Face Hub.
+
+```bash
+export MODEL_DIR="stabilityai/stable-diffusion-xl-base-1.0"
+export OUTPUT_DIR="path to save model"
+
+accelerate launch train_controlnet_sdxl.py \
+ --pretrained_model_name_or_path=$MODEL_DIR \
+ --output_dir=$OUTPUT_DIR \
+ --dataset_name=fusing/fill50k \
+ --mixed_precision="fp16" \
+ --resolution=1024 \
+ --learning_rate=1e-5 \
+ --max_train_steps=15000 \
+ --validation_image "./conditioning_image_1.png" "./conditioning_image_2.png" \
+ --validation_prompt "red circle with blue background" "cyan circle with brown floral background" \
+ --validation_steps=100 \
+ --train_batch_size=1 \
+ --gradient_accumulation_steps=4 \
+ --report_to="wandb" \
+ --seed=42 \
+ --push_to_hub
+```
+
+To better track our training experiments, we're using the following flags in the command above:
+
+* `report_to="wandb` will ensure the training runs are tracked on Weights and Biases. To use it, be sure to install `wandb` with `pip install wandb`.
+* `validation_image`, `validation_prompt`, and `validation_steps` to allow the script to do a few validation inference runs. This allows us to qualitatively check if the training is progressing as expected.
+
+Our experiments were conducted on a single 40GB A100 GPU.
+
+### Inference
+
+Once training is done, we can perform inference like so:
+
+```python
+from diffusers import StableDiffusionXLControlNetPipeline, ControlNetModel, UniPCMultistepScheduler
+from diffusers.utils import load_image
+import torch
+
+base_model_path = "stabilityai/stable-diffusion-xl-base-1.0"
+controlnet_path = "path to controlnet"
+
+controlnet = ControlNetModel.from_pretrained(controlnet_path, torch_dtype=torch.float16)
+pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
+    base_model_path, controlnet=controlnet, torch_dtype=torch.float16
+)
+
+# speed up diffusion process with faster scheduler and memory optimization
+pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config)
+# remove following line if xformers is not installed or when using Torch 2.0.
+pipe.enable_xformers_memory_efficient_attention()
+# memory optimization.
+pipe.enable_model_cpu_offload()
+
+control_image = load_image("./conditioning_image_1.png").resize((1024, 1024))
+prompt = "pale golden rod circle with old lace background"
+
+# generate image
+generator = torch.manual_seed(0)
+image = pipe(
+    prompt, num_inference_steps=20, generator=generator, image=control_image
+).images[0]
+image.save("./output.png")
+```
+
+## Notes
+
+### Specifying a better VAE
+
+SDXL's VAE is known to suffer from numerical instability issues. This is why we also expose a CLI argument namely `--pretrained_vae_model_name_or_path` that lets you specify the location of an alternative VAE (such as [`madebyollin/sdxl-vae-fp16-fix`](https://huggingface.co/madebyollin/sdxl-vae-fp16-fix)).
+
+If you're using this VAE during training, you need to ensure you're using it during inference too. You do so by:
+
+```diff
++ vae = AutoencoderKL.from_pretrained(vae_path_or_repo_id, torch_dtype=torch.float16)
+controlnet = ControlNetModel.from_pretrained(controlnet_path, torch_dtype=torch.float16)
+pipe = StableDiffusionXLControlNetPipeline.from_pretrained(
+    base_model_path, controlnet=controlnet, torch_dtype=torch.float16,
++   vae=vae,
+)

config.json

@@ -0,0 +1,52 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.33.0.dev0",
+  "_name_or_path": "lllyasviel/sd-controlnet-canny",
+  "act_fn": "silu",
+  "addition_embed_type": null,
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": null,
+  "attention_head_dim": 8,
+  "block_out_channels": [
+    320,
+    640,
+    1280,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 768,
+  "down_block_types": [
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D",
+    "DownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": null,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": 1,
+  "upcast_attention": false,
+  "use_linear_projection": false
+}

diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b554b86ce5bfdda631f6aa02875fe5cfeabbb019b1de2eddaaf5f939cc9191b
+size 1445157120

logs/controlnet-2210/1743377945.961673/events.out.tfevents.1743377945.79c7fdbf50b7.8420.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:389766e60dd13e2b0eed03964d3ab8b8bcc8cd678fa14134ce4467d466684b9c
+size 2436

logs/controlnet-2210/1743377945.9631884/hparams.yml

@@ -0,0 +1,50 @@
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+caption_column: text
+checkpointing_steps: 500
+checkpoints_total_limit: null
+conditioning_image_column: conditioning_image
+controlnet_model_name_or_path: lllyasviel/sd-controlnet-canny
+dataloader_num_workers: 0
+dataset_config_name: null
+dataset_name: brity01/csc2210_1
+enable_xformers_memory_efficient_attention: false
+gradient_accumulation_steps: 4
+gradient_checkpointing: false
+hub_model_id: controlnet-2210
+hub_token: null
+image_column: image
+learning_rate: 1.0e-05
+logging_dir: logs
+lr_num_cycles: 1
+lr_power: 1.0
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 62
+mixed_precision: null
+num_train_epochs: 1
+num_validation_images: 4
+output_dir: ./
+pretrained_model_name_or_path: stable-diffusion-v1-5/stable-diffusion-v1-5
+proportion_empty_prompts: 0
+push_to_hub: true
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: null
+set_grads_to_none: false
+tokenizer_name: null
+tracker_project_name: controlnet-2210
+train_batch_size: 1
+train_data_dir: null
+use_8bit_adam: false
+validation_steps: 100
+variant: null

logs/controlnet-2210/events.out.tfevents.1743377945.79c7fdbf50b7.8420.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:108cf9cd114aeee66f4524efbc9c1053764fcf1dbf2dd2c7d6ae5213888357e6
+size 4185798

logs/train_controlnet/1743377496.261348/events.out.tfevents.1743377496.79c7fdbf50b7.6238.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:769b6306a2c6336554efc36cc87711fdc972836b04518f5a44ddcfa2926bd122
+size 2384

logs/train_controlnet/1743377496.2628558/hparams.yml

@@ -0,0 +1,50 @@
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+caption_column: text
+checkpointing_steps: 500
+checkpoints_total_limit: null
+conditioning_image_column: conditioning_image
+controlnet_model_name_or_path: lllyasviel/sd-controlnet-canny
+dataloader_num_workers: 0
+dataset_config_name: null
+dataset_name: brity01/csc2210_1
+enable_xformers_memory_efficient_attention: false
+gradient_accumulation_steps: 4
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+image_column: image
+learning_rate: 1.0e-05
+logging_dir: logs
+lr_num_cycles: 1
+lr_power: 1.0
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 62
+mixed_precision: null
+num_train_epochs: 1
+num_validation_images: 4
+output_dir: ./
+pretrained_model_name_or_path: stable-diffusion-v1-5/stable-diffusion-v1-5
+proportion_empty_prompts: 0
+push_to_hub: false
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: null
+set_grads_to_none: false
+tokenizer_name: null
+tracker_project_name: train_controlnet
+train_batch_size: 1
+train_data_dir: null
+use_8bit_adam: false
+validation_steps: 100
+variant: null

logs/train_controlnet/events.out.tfevents.1743377496.79c7fdbf50b7.6238.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f84680b134f5cea5fe92681258df13a8679fc4b10cabf0dc998ff68d47c0c89
+size 4251047

requirements.txt

@@ -0,0 +1,6 @@
+accelerate>=0.16.0
+torchvision
+transformers>=4.25.1
+ftfy
+tensorboard
+datasets

requirements_flax.txt

@@ -0,0 +1,9 @@
+transformers>=4.25.1
+datasets
+flax
+optax
+torch
+torchvision
+ftfy
+tensorboard
+Jinja2

requirements_flux.txt

@@ -0,0 +1,9 @@
+accelerate>=0.16.0
+torchvision
+transformers>=4.25.1
+ftfy
+tensorboard
+Jinja2
+datasets
+wandb
+SentencePiece

requirements_sd3.txt

@@ -0,0 +1,8 @@
+accelerate>=0.16.0
+torchvision
+transformers>=4.25.1
+ftfy
+tensorboard
+Jinja2
+datasets
+wandb

requirements_sdxl.txt

@@ -0,0 +1,8 @@
+accelerate>=0.16.0
+torchvision
+transformers>=4.25.1
+ftfy
+tensorboard
+Jinja2
+datasets
+wandb

test_controlnet.py

@@ -0,0 +1,184 @@
+# coding=utf-8
+# Copyright 2024 HuggingFace Inc.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+import logging
+import os
+import sys
+import tempfile
+
+
+sys.path.append("..")
+from test_examples_utils import ExamplesTestsAccelerate, run_command  # noqa: E402
+
+
+logging.basicConfig(level=logging.DEBUG)
+
+logger = logging.getLogger()
+stream_handler = logging.StreamHandler(sys.stdout)
+logger.addHandler(stream_handler)
+
+
+class ControlNet(ExamplesTestsAccelerate):
+    def test_controlnet_checkpointing_checkpoints_total_limit(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-stable-diffusion-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --max_train_steps=6
+            --checkpoints_total_limit=2
+            --checkpointing_steps=2
+            --controlnet_model_name_or_path=hf-internal-testing/tiny-controlnet
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertEqual(
+                {x for x in os.listdir(tmpdir) if "checkpoint" in x},
+                {"checkpoint-4", "checkpoint-6"},
+            )
+
+    def test_controlnet_checkpointing_checkpoints_total_limit_removes_multiple_checkpoints(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-stable-diffusion-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --controlnet_model_name_or_path=hf-internal-testing/tiny-controlnet
+            --max_train_steps=6
+            --checkpointing_steps=2
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertEqual(
+                {x for x in os.listdir(tmpdir) if "checkpoint" in x},
+                {"checkpoint-2", "checkpoint-4", "checkpoint-6"},
+            )
+
+            resume_run_args = f"""
+            examples/controlnet/train_controlnet.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-stable-diffusion-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --controlnet_model_name_or_path=hf-internal-testing/tiny-controlnet
+            --max_train_steps=8
+            --checkpointing_steps=2
+            --resume_from_checkpoint=checkpoint-6
+            --checkpoints_total_limit=2
+            """.split()
+
+            run_command(self._launch_args + resume_run_args)
+
+            self.assertEqual({x for x in os.listdir(tmpdir) if "checkpoint" in x}, {"checkpoint-6", "checkpoint-8"})
+
+
+class ControlNetSDXL(ExamplesTestsAccelerate):
+    def test_controlnet_sdxl(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet_sdxl.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-stable-diffusion-xl-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --controlnet_model_name_or_path=hf-internal-testing/tiny-controlnet-sdxl
+            --max_train_steps=4
+            --checkpointing_steps=2
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "diffusion_pytorch_model.safetensors")))
+
+
+class ControlNetSD3(ExamplesTestsAccelerate):
+    def test_controlnet_sd3(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet_sd3.py
+            --pretrained_model_name_or_path=DavyMorgan/tiny-sd3-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --controlnet_model_name_or_path=DavyMorgan/tiny-controlnet-sd3
+            --max_train_steps=4
+            --checkpointing_steps=2
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "diffusion_pytorch_model.safetensors")))
+
+
+class ControlNetSD35(ExamplesTestsAccelerate):
+    def test_controlnet_sd3(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet_sd3.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-sd35-pipe
+            --dataset_name=hf-internal-testing/fill10
+            --output_dir={tmpdir}
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --controlnet_model_name_or_path=DavyMorgan/tiny-controlnet-sd35
+            --max_train_steps=4
+            --checkpointing_steps=2
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "diffusion_pytorch_model.safetensors")))
+
+
+class ControlNetflux(ExamplesTestsAccelerate):
+    def test_controlnet_flux(self):
+        with tempfile.TemporaryDirectory() as tmpdir:
+            test_args = f"""
+            examples/controlnet/train_controlnet_flux.py
+            --pretrained_model_name_or_path=hf-internal-testing/tiny-flux-pipe
+            --output_dir={tmpdir}
+            --dataset_name=hf-internal-testing/fill10
+            --conditioning_image_column=conditioning_image
+            --image_column=image
+            --caption_column=text
+            --resolution=64
+            --train_batch_size=1
+            --gradient_accumulation_steps=1
+            --max_train_steps=4
+            --checkpointing_steps=2
+            --num_double_layers=1
+            --num_single_layers=1
+            """.split()
+
+            run_command(self._launch_args + test_args)
+
+            self.assertTrue(os.path.isfile(os.path.join(tmpdir, "diffusion_pytorch_model.safetensors")))

train_controlnet.py

@@ -0,0 +1,1185 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+import argparse
+import contextlib
+import gc
+import logging
+import math
+import os
+import random
+import shutil
+from pathlib import Path
+
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, PretrainedConfig
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    ControlNetModel,
+    DDPMScheduler,
+    StableDiffusionControlNetPipeline,
+    UNet2DConditionModel,
+    UniPCMultistepScheduler,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+from diffusers.utils.import_utils import is_xformers_available
+from diffusers.utils.torch_utils import is_compiled_module
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.33.0.dev0")
+
+logger = get_logger(__name__)
+
+
+def image_grid(imgs, rows, cols):
+    assert len(imgs) == rows * cols
+
+    w, h = imgs[0].size
+    grid = Image.new("RGB", size=(cols * w, rows * h))
+
+    for i, img in enumerate(imgs):
+        grid.paste(img, box=(i % cols * w, i // cols * h))
+    return grid
+
+
+def log_validation(
+    vae, text_encoder, tokenizer, unet, controlnet, args, accelerator, weight_dtype, step, is_final_validation=False
+):
+    logger.info("Running validation... ")
+
+    if not is_final_validation:
+        controlnet = accelerator.unwrap_model(controlnet)
+    else:
+        controlnet = ControlNetModel.from_pretrained(args.output_dir, torch_dtype=weight_dtype)
+
+    pipeline = StableDiffusionControlNetPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        vae=vae,
+        text_encoder=text_encoder,
+        tokenizer=tokenizer,
+        unet=unet,
+        controlnet=controlnet,
+        safety_checker=None,
+        revision=args.revision,
+        variant=args.variant,
+        torch_dtype=weight_dtype,
+    )
+    pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+    inference_ctx = contextlib.nullcontext() if is_final_validation else torch.autocast("cuda")
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        validation_image = Image.open(validation_image).convert("RGB")
+
+        images = []
+
+        for _ in range(args.num_validation_images):
+            with inference_ctx:
+                image = pipeline(
+                    validation_prompt, validation_image, num_inference_steps=20, generator=generator
+                ).images[0]
+
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    tracker_key = "test" if is_final_validation else "validation"
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = [np.asarray(validation_image)]
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({tracker_key: formatted_images})
+        else:
+            logger.warning(f"image logging not implemented for {tracker.name}")
+
+        del pipeline
+        gc.collect()
+        torch.cuda.empty_cache()
+
+        return image_logs
+
+
+def import_model_class_from_model_name_or_path(pretrained_model_name_or_path: str, revision: str):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path,
+        subfolder="text_encoder",
+        revision=revision,
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "RobertaSeriesModelWithTransformation":
+        from diffusers.pipelines.alt_diffusion.modeling_roberta_series import RobertaSeriesModelWithTransformation
+
+        return RobertaSeriesModelWithTransformation
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+"""
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="creativeml-openrail-m",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "stable-diffusion",
+        "stable-diffusion-diffusers",
+        "text-to-image",
+        "diffusers",
+        "controlnet",
+        "diffusers-training",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--variant",
+        type=str,
+        default=None,
+        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def make_train_dataset(args, tokenizer, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+            data_dir=args.train_data_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    def tokenize_captions(examples, is_train=True):
+        captions = []
+        for caption in examples[caption_column]:
+            if random.random() < args.proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+            else:
+                raise ValueError(
+                    f"Caption column `{caption_column}` should contain either strings or lists of strings."
+                )
+        inputs = tokenizer(
+            captions, max_length=tokenizer.model_max_length, padding="max_length", truncation=True, return_tensors="pt"
+        )
+        return inputs.input_ids
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [image.convert("RGB") for image in examples[image_column]]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [image.convert("RGB") for image in examples[conditioning_image_column]]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+        examples["input_ids"] = tokenize_captions(examples)
+
+        return examples
+
+    with accelerator.main_process_first():
+        if args.max_train_samples is not None:
+            dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
+        # Set the training transforms
+        train_dataset = dataset["train"].with_transform(preprocess_train)
+
+    return train_dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.stack([example["input_ids"] for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "input_ids": input_ids,
+    }
+
+
+def main(args):
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Disable AMP for MPS.
+    if torch.backends.mps.is_available():
+        accelerator.native_amp = False
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer
+    if args.tokenizer_name:
+        tokenizer = AutoTokenizer.from_pretrained(args.tokenizer_name, revision=args.revision, use_fast=False)
+    elif args.pretrained_model_name_or_path:
+        tokenizer = AutoTokenizer.from_pretrained(
+            args.pretrained_model_name_or_path,
+            subfolder="tokenizer",
+            revision=args.revision,
+            use_fast=False,
+        )
+
+    # import correct text encoder class
+    text_encoder_cls = import_model_class_from_model_name_or_path(args.pretrained_model_name_or_path, args.revision)
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder = text_encoder_cls.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    vae = AutoencoderKL.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="vae", revision=args.revision, variant=args.variant
+    )
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision, variant=args.variant
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        controlnet = ControlNetModel.from_unet(unet)
+
+    # Taken from [Sayak Paul's Diffusers PR #6511](https://github.com/huggingface/diffusers/pull/6511/files)
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            if accelerator.is_main_process:
+                i = len(weights) - 1
+
+                while len(weights) > 0:
+                    weights.pop()
+                    model = models[i]
+
+                    sub_dir = "controlnet"
+                    model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                    i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    vae.requires_grad_(False)
+    unet.requires_grad_(False)
+    text_encoder.requires_grad_(False)
+    controlnet.train()
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warning(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+            controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    train_dataset = make_train_dataset(args, tokenizer, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
+        num_training_steps=args.max_train_steps * accelerator.num_processes,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, unet and text_encoder to device and cast to weight_dtype
+    vae.to(accelerator.device, dtype=weight_dtype)
+    unet.to(accelerator.device, dtype=weight_dtype)
+    text_encoder.to(accelerator.device, dtype=weight_dtype)
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                latents = vae.encode(batch["pixel_values"].to(dtype=weight_dtype)).latent_dist.sample()
+                latents = latents * vae.config.scaling_factor
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+                # Sample a random timestep for each image
+                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents.float(), noise.float(), timesteps).to(
+                    dtype=weight_dtype
+                )
+
+                # Get the text embedding for conditioning
+                encoder_hidden_states = text_encoder(batch["input_ids"], return_dict=False)[0]
+
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+
+                down_block_res_samples, mid_block_res_sample = controlnet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=encoder_hidden_states,
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )
+
+                # Predict the noise residual
+                model_pred = unet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=encoder_hidden_states,
+                    down_block_additional_residuals=[
+                        sample.to(dtype=weight_dtype) for sample in down_block_res_samples
+                    ],
+                    mid_block_additional_residual=mid_block_res_sample.to(dtype=weight_dtype),
+                    return_dict=False,
+                )[0]
+
+                # Get the target for loss depending on the prediction type
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    target = noise
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                else:
+                    raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+                loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                if accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae,
+                            text_encoder,
+                            tokenizer,
+                            unet,
+                            controlnet,
+                            args,
+                            accelerator,
+                            weight_dtype,
+                            global_step,
+                        )
+
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        # Run a final round of validation.
+        image_logs = None
+        if args.validation_prompt is not None:
+            image_logs = log_validation(
+                vae=vae,
+                text_encoder=text_encoder,
+                tokenizer=tokenizer,
+                unet=unet,
+                controlnet=None,
+                args=args,
+                accelerator=accelerator,
+                weight_dtype=weight_dtype,
+                step=global_step,
+                is_final_validation=True,
+            )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

train_controlnet_flax.py

@@ -0,0 +1,1152 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+import argparse
+import logging
+import math
+import os
+import random
+import time
+from pathlib import Path
+
+import jax
+import jax.numpy as jnp
+import numpy as np
+import optax
+import torch
+import torch.utils.checkpoint
+import transformers
+from datasets import load_dataset, load_from_disk
+from flax import jax_utils
+from flax.core.frozen_dict import unfreeze
+from flax.training import train_state
+from flax.training.common_utils import shard
+from huggingface_hub import create_repo, upload_folder
+from PIL import Image, PngImagePlugin
+from torch.utils.data import IterableDataset
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import CLIPTokenizer, FlaxCLIPTextModel, set_seed
+
+from diffusers import (
+    FlaxAutoencoderKL,
+    FlaxControlNetModel,
+    FlaxDDPMScheduler,
+    FlaxStableDiffusionControlNetPipeline,
+    FlaxUNet2DConditionModel,
+)
+from diffusers.utils import check_min_version, is_wandb_available, make_image_grid
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+
+
+# To prevent an error that occurs when there are abnormally large compressed data chunk in the png image
+# see more https://github.com/python-pillow/Pillow/issues/5610
+LARGE_ENOUGH_NUMBER = 100
+PngImagePlugin.MAX_TEXT_CHUNK = LARGE_ENOUGH_NUMBER * (1024**2)
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.33.0.dev0")
+
+logger = logging.getLogger(__name__)
+
+
+def log_validation(pipeline, pipeline_params, controlnet_params, tokenizer, args, rng, weight_dtype):
+    logger.info("Running validation...")
+
+    pipeline_params = pipeline_params.copy()
+    pipeline_params["controlnet"] = controlnet_params
+
+    num_samples = jax.device_count()
+    prng_seed = jax.random.split(rng, jax.device_count())
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        prompts = num_samples * [validation_prompt]
+        prompt_ids = pipeline.prepare_text_inputs(prompts)
+        prompt_ids = shard(prompt_ids)
+
+        validation_image = Image.open(validation_image).convert("RGB")
+        processed_image = pipeline.prepare_image_inputs(num_samples * [validation_image])
+        processed_image = shard(processed_image)
+        images = pipeline(
+            prompt_ids=prompt_ids,
+            image=processed_image,
+            params=pipeline_params,
+            prng_seed=prng_seed,
+            num_inference_steps=50,
+            jit=True,
+        ).images
+
+        images = images.reshape((images.shape[0] * images.shape[1],) + images.shape[-3:])
+        images = pipeline.numpy_to_pil(images)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    if args.report_to == "wandb":
+        formatted_images = []
+        for log in image_logs:
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+
+            formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+            for image in images:
+                image = wandb.Image(image, caption=validation_prompt)
+                formatted_images.append(image)
+
+        wandb.log({"validation": formatted_images})
+    else:
+        logger.warning(f"image logging not implemented for {args.report_to}")
+
+    return image_logs
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            make_image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# controlnet- {repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning. You can find some example images in the following. \n
+{img_str}
+"""
+
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="creativeml-openrail-m",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "stable-diffusion",
+        "stable-diffusion-diffusers",
+        "text-to-image",
+        "diffusers",
+        "controlnet",
+        "jax-diffusers-event",
+        "diffusers-training",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args():
+    parser = argparse.ArgumentParser(description="Simple example of a training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--from_pt",
+        action="store_true",
+        help="Load the pretrained model from a PyTorch checkpoint.",
+    )
+    parser.add_argument(
+        "--controlnet_revision",
+        type=str,
+        default=None,
+        help="Revision of controlnet model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--profile_steps",
+        type=int,
+        default=0,
+        help="How many training steps to profile in the beginning.",
+    )
+    parser.add_argument(
+        "--profile_validation",
+        action="store_true",
+        help="Whether to profile the (last) validation.",
+    )
+    parser.add_argument(
+        "--profile_memory",
+        action="store_true",
+        help="Whether to dump an initial (before training loop) and a final (at program end) memory profile.",
+    )
+    parser.add_argument(
+        "--ccache",
+        type=str,
+        default=None,
+        help="Enables compilation cache.",
+    )
+    parser.add_argument(
+        "--controlnet_from_pt",
+        action="store_true",
+        help="Load the controlnet model from a PyTorch checkpoint.",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="runs/{timestamp}",
+        help="The output directory where the model predictions and checkpoints will be written. "
+        "Can contain placeholders: {timestamp}.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=0, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=1, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=100)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=5000,
+        help=("Save a checkpoint of the training state every X updates."),
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=1e-4,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--snr_gamma",
+        type=float,
+        default=None,
+        help="SNR weighting gamma to be used if rebalancing the loss. Recommended value is 5.0. "
+        "More details here: https://arxiv.org/abs/2303.09556.",
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_steps",
+        type=int,
+        default=100,
+        help=("log training metric every X steps to `--report_t`"),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="wandb",
+        help=('The integration to report the results and logs to. Currently only supported platforms are `"wandb"`'),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default="no",
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose"
+            "between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >= 1.10."
+            "and an Nvidia Ampere GPU."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument("--streaming", action="store_true", help="To stream a large dataset from Hub.")
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training dataset. By default it will use `load_dataset` method to load a custom dataset from the folder."
+            "Folder must contain a dataset script as described here https://huggingface.co/docs/datasets/dataset_script) ."
+            "If `--load_from_disk` flag is passed, it will use `load_from_disk` method instead. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--load_from_disk",
+        action="store_true",
+        help=(
+            "If True, will load a dataset that was previously saved using `save_to_disk` from `--train_data_dir`"
+            "See more https://huggingface.co/docs/datasets/package_reference/main_classes#datasets.Dataset.load_from_disk"
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set. Needed if `streaming` is set to True."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` and logging the images."
+        ),
+    )
+    parser.add_argument("--wandb_entity", type=str, default=None, help=("The wandb entity to use (for teams)."))
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="train_controlnet_flax",
+        help=("The `project` argument passed to wandb"),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps", type=int, default=1, help="Number of steps to accumulate gradients over"
+    )
+    parser.add_argument("--local_rank", type=int, default=-1, help="For distributed training: local_rank")
+
+    args = parser.parse_args()
+    args.output_dir = args.output_dir.replace("{timestamp}", time.strftime("%Y%m%d_%H%M%S"))
+
+    env_local_rank = int(os.environ.get("LOCAL_RANK", -1))
+    if env_local_rank != -1 and env_local_rank != args.local_rank:
+        args.local_rank = env_local_rank
+
+    # Sanity checks
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Need either a dataset name or a training folder.")
+    if args.dataset_name is not None and args.train_data_dir is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    # This idea comes from
+    # https://github.com/borisdayma/dalle-mini/blob/d2be512d4a6a9cda2d63ba04afc33038f98f705f/src/dalle_mini/data.py#L370
+    if args.streaming and args.max_train_samples is None:
+        raise ValueError("You must specify `max_train_samples` when using dataset streaming.")
+
+    return args
+
+
+def make_train_dataset(args, tokenizer, batch_size=None):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+            streaming=args.streaming,
+        )
+    else:
+        if args.train_data_dir is not None:
+            if args.load_from_disk:
+                dataset = load_from_disk(
+                    args.train_data_dir,
+                )
+            else:
+                dataset = load_dataset(
+                    args.train_data_dir,
+                    cache_dir=args.cache_dir,
+                )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    if isinstance(dataset["train"], IterableDataset):
+        column_names = next(iter(dataset["train"])).keys()
+    else:
+        column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {caption_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    def tokenize_captions(examples, is_train=True):
+        captions = []
+        for caption in examples[caption_column]:
+            if random.random() < args.proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+            else:
+                raise ValueError(
+                    f"Caption column `{caption_column}` should contain either strings or lists of strings."
+                )
+        inputs = tokenizer(
+            captions, max_length=tokenizer.model_max_length, padding="max_length", truncation=True, return_tensors="pt"
+        )
+        return inputs.input_ids
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [image.convert("RGB") for image in examples[image_column]]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [image.convert("RGB") for image in examples[conditioning_image_column]]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+        examples["input_ids"] = tokenize_captions(examples)
+
+        return examples
+
+    if jax.process_index() == 0:
+        if args.max_train_samples is not None:
+            if args.streaming:
+                dataset["train"] = dataset["train"].shuffle(seed=args.seed).take(args.max_train_samples)
+            else:
+                dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
+        # Set the training transforms
+        if args.streaming:
+            train_dataset = dataset["train"].map(
+                preprocess_train,
+                batched=True,
+                batch_size=batch_size,
+                remove_columns=list(dataset["train"].features.keys()),
+            )
+        else:
+            train_dataset = dataset["train"].with_transform(preprocess_train)
+
+    return train_dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    input_ids = torch.stack([example["input_ids"] for example in examples])
+
+    batch = {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "input_ids": input_ids,
+    }
+    batch = {k: v.numpy() for k, v in batch.items()}
+    return batch
+
+
+def get_params_to_save(params):
+    return jax.device_get(jax.tree_util.tree_map(lambda x: x[0], params))
+
+
+def main():
+    args = parse_args()
+
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    # Setup logging, we only want one process per machine to log things on the screen.
+    logger.setLevel(logging.INFO if jax.process_index() == 0 else logging.ERROR)
+    if jax.process_index() == 0:
+        transformers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+
+    # wandb init
+    if jax.process_index() == 0 and args.report_to == "wandb":
+        wandb.init(
+            entity=args.wandb_entity,
+            project=args.tracker_project_name,
+            job_type="train",
+            config=args,
+        )
+
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    rng = jax.random.PRNGKey(0)
+
+    # Handle the repository creation
+    if jax.process_index() == 0:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer and add the placeholder token as a additional special token
+    if args.tokenizer_name:
+        tokenizer = CLIPTokenizer.from_pretrained(args.tokenizer_name)
+    elif args.pretrained_model_name_or_path:
+        tokenizer = CLIPTokenizer.from_pretrained(
+            args.pretrained_model_name_or_path, subfolder="tokenizer", revision=args.revision
+        )
+    else:
+        raise NotImplementedError("No tokenizer specified!")
+
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    total_train_batch_size = args.train_batch_size * jax.local_device_count() * args.gradient_accumulation_steps
+    train_dataset = make_train_dataset(args, tokenizer, batch_size=total_train_batch_size)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=not args.streaming,
+        collate_fn=collate_fn,
+        batch_size=total_train_batch_size,
+        num_workers=args.dataloader_num_workers,
+        drop_last=True,
+    )
+
+    weight_dtype = jnp.float32
+    if args.mixed_precision == "fp16":
+        weight_dtype = jnp.float16
+    elif args.mixed_precision == "bf16":
+        weight_dtype = jnp.bfloat16
+
+    # Load models and create wrapper for stable diffusion
+    text_encoder = FlaxCLIPTextModel.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="text_encoder",
+        dtype=weight_dtype,
+        revision=args.revision,
+        from_pt=args.from_pt,
+    )
+    vae, vae_params = FlaxAutoencoderKL.from_pretrained(
+        args.pretrained_model_name_or_path,
+        revision=args.revision,
+        subfolder="vae",
+        dtype=weight_dtype,
+        from_pt=args.from_pt,
+    )
+    unet, unet_params = FlaxUNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="unet",
+        dtype=weight_dtype,
+        revision=args.revision,
+        from_pt=args.from_pt,
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet, controlnet_params = FlaxControlNetModel.from_pretrained(
+            args.controlnet_model_name_or_path,
+            revision=args.controlnet_revision,
+            from_pt=args.controlnet_from_pt,
+            dtype=jnp.float32,
+        )
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        rng, rng_params = jax.random.split(rng)
+
+        controlnet = FlaxControlNetModel(
+            in_channels=unet.config.in_channels,
+            down_block_types=unet.config.down_block_types,
+            only_cross_attention=unet.config.only_cross_attention,
+            block_out_channels=unet.config.block_out_channels,
+            layers_per_block=unet.config.layers_per_block,
+            attention_head_dim=unet.config.attention_head_dim,
+            cross_attention_dim=unet.config.cross_attention_dim,
+            use_linear_projection=unet.config.use_linear_projection,
+            flip_sin_to_cos=unet.config.flip_sin_to_cos,
+            freq_shift=unet.config.freq_shift,
+        )
+        controlnet_params = controlnet.init_weights(rng=rng_params)
+        controlnet_params = unfreeze(controlnet_params)
+        for key in [
+            "conv_in",
+            "time_embedding",
+            "down_blocks_0",
+            "down_blocks_1",
+            "down_blocks_2",
+            "down_blocks_3",
+            "mid_block",
+        ]:
+            controlnet_params[key] = unet_params[key]
+
+    pipeline, pipeline_params = FlaxStableDiffusionControlNetPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        tokenizer=tokenizer,
+        controlnet=controlnet,
+        safety_checker=None,
+        dtype=weight_dtype,
+        revision=args.revision,
+        from_pt=args.from_pt,
+    )
+    pipeline_params = jax_utils.replicate(pipeline_params)
+
+    # Optimization
+    if args.scale_lr:
+        args.learning_rate = args.learning_rate * total_train_batch_size
+
+    constant_scheduler = optax.constant_schedule(args.learning_rate)
+
+    adamw = optax.adamw(
+        learning_rate=constant_scheduler,
+        b1=args.adam_beta1,
+        b2=args.adam_beta2,
+        eps=args.adam_epsilon,
+        weight_decay=args.adam_weight_decay,
+    )
+
+    optimizer = optax.chain(
+        optax.clip_by_global_norm(args.max_grad_norm),
+        adamw,
+    )
+
+    state = train_state.TrainState.create(apply_fn=controlnet.__call__, params=controlnet_params, tx=optimizer)
+
+    noise_scheduler, noise_scheduler_state = FlaxDDPMScheduler.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="scheduler"
+    )
+
+    # Initialize our training
+    validation_rng, train_rngs = jax.random.split(rng)
+    train_rngs = jax.random.split(train_rngs, jax.local_device_count())
+
+    def compute_snr(timesteps):
+        """
+        Computes SNR as per https://github.com/TiankaiHang/Min-SNR-Diffusion-Training/blob/521b624bd70c67cee4bdf49225915f5945a872e3/guided_diffusion/gaussian_diffusion.py#L847-L849
+        """
+        alphas_cumprod = noise_scheduler_state.common.alphas_cumprod
+        sqrt_alphas_cumprod = alphas_cumprod**0.5
+        sqrt_one_minus_alphas_cumprod = (1.0 - alphas_cumprod) ** 0.5
+
+        alpha = sqrt_alphas_cumprod[timesteps]
+        sigma = sqrt_one_minus_alphas_cumprod[timesteps]
+        # Compute SNR.
+        snr = (alpha / sigma) ** 2
+        return snr
+
+    def train_step(state, unet_params, text_encoder_params, vae_params, batch, train_rng):
+        # reshape batch, add grad_step_dim if gradient_accumulation_steps > 1
+        if args.gradient_accumulation_steps > 1:
+            grad_steps = args.gradient_accumulation_steps
+            batch = jax.tree_map(lambda x: x.reshape((grad_steps, x.shape[0] // grad_steps) + x.shape[1:]), batch)
+
+        def compute_loss(params, minibatch, sample_rng):
+            # Convert images to latent space
+            vae_outputs = vae.apply(
+                {"params": vae_params}, minibatch["pixel_values"], deterministic=True, method=vae.encode
+            )
+            latents = vae_outputs.latent_dist.sample(sample_rng)
+            # (NHWC) -> (NCHW)
+            latents = jnp.transpose(latents, (0, 3, 1, 2))
+            latents = latents * vae.config.scaling_factor
+
+            # Sample noise that we'll add to the latents
+            noise_rng, timestep_rng = jax.random.split(sample_rng)
+            noise = jax.random.normal(noise_rng, latents.shape)
+            # Sample a random timestep for each image
+            bsz = latents.shape[0]
+            timesteps = jax.random.randint(
+                timestep_rng,
+                (bsz,),
+                0,
+                noise_scheduler.config.num_train_timesteps,
+            )
+
+            # Add noise to the latents according to the noise magnitude at each timestep
+            # (this is the forward diffusion process)
+            noisy_latents = noise_scheduler.add_noise(noise_scheduler_state, latents, noise, timesteps)
+
+            # Get the text embedding for conditioning
+            encoder_hidden_states = text_encoder(
+                minibatch["input_ids"],
+                params=text_encoder_params,
+                train=False,
+            )[0]
+
+            controlnet_cond = minibatch["conditioning_pixel_values"]
+
+            # Predict the noise residual and compute loss
+            down_block_res_samples, mid_block_res_sample = controlnet.apply(
+                {"params": params},
+                noisy_latents,
+                timesteps,
+                encoder_hidden_states,
+                controlnet_cond,
+                train=True,
+                return_dict=False,
+            )
+
+            model_pred = unet.apply(
+                {"params": unet_params},
+                noisy_latents,
+                timesteps,
+                encoder_hidden_states,
+                down_block_additional_residuals=down_block_res_samples,
+                mid_block_additional_residual=mid_block_res_sample,
+            ).sample
+
+            # Get the target for loss depending on the prediction type
+            if noise_scheduler.config.prediction_type == "epsilon":
+                target = noise
+            elif noise_scheduler.config.prediction_type == "v_prediction":
+                target = noise_scheduler.get_velocity(noise_scheduler_state, latents, noise, timesteps)
+            else:
+                raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+
+            loss = (target - model_pred) ** 2
+
+            if args.snr_gamma is not None:
+                snr = jnp.array(compute_snr(timesteps))
+                snr_loss_weights = jnp.where(snr < args.snr_gamma, snr, jnp.ones_like(snr) * args.snr_gamma)
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    snr_loss_weights = snr_loss_weights / snr
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    snr_loss_weights = snr_loss_weights / (snr + 1)
+
+                loss = loss * snr_loss_weights
+
+            loss = loss.mean()
+
+            return loss
+
+        grad_fn = jax.value_and_grad(compute_loss)
+
+        # get a minibatch (one gradient accumulation slice)
+        def get_minibatch(batch, grad_idx):
+            return jax.tree_util.tree_map(
+                lambda x: jax.lax.dynamic_index_in_dim(x, grad_idx, keepdims=False),
+                batch,
+            )
+
+        def loss_and_grad(grad_idx, train_rng):
+            # create minibatch for the grad step
+            minibatch = get_minibatch(batch, grad_idx) if grad_idx is not None else batch
+            sample_rng, train_rng = jax.random.split(train_rng, 2)
+            loss, grad = grad_fn(state.params, minibatch, sample_rng)
+            return loss, grad, train_rng
+
+        if args.gradient_accumulation_steps == 1:
+            loss, grad, new_train_rng = loss_and_grad(None, train_rng)
+        else:
+            init_loss_grad_rng = (
+                0.0,  # initial value for cumul_loss
+                jax.tree_map(jnp.zeros_like, state.params),  # initial value for cumul_grad
+                train_rng,  # initial value for train_rng
+            )
+
+            def cumul_grad_step(grad_idx, loss_grad_rng):
+                cumul_loss, cumul_grad, train_rng = loss_grad_rng
+                loss, grad, new_train_rng = loss_and_grad(grad_idx, train_rng)
+                cumul_loss, cumul_grad = jax.tree_map(jnp.add, (cumul_loss, cumul_grad), (loss, grad))
+                return cumul_loss, cumul_grad, new_train_rng
+
+            loss, grad, new_train_rng = jax.lax.fori_loop(
+                0,
+                args.gradient_accumulation_steps,
+                cumul_grad_step,
+                init_loss_grad_rng,
+            )
+            loss, grad = jax.tree_map(lambda x: x / args.gradient_accumulation_steps, (loss, grad))
+
+        grad = jax.lax.pmean(grad, "batch")
+
+        new_state = state.apply_gradients(grads=grad)
+
+        metrics = {"loss": loss}
+        metrics = jax.lax.pmean(metrics, axis_name="batch")
+
+        def l2(xs):
+            return jnp.sqrt(sum([jnp.vdot(x, x) for x in jax.tree_util.tree_leaves(xs)]))
+
+        metrics["l2_grads"] = l2(jax.tree_util.tree_leaves(grad))
+
+        return new_state, metrics, new_train_rng
+
+    # Create parallel version of the train step
+    p_train_step = jax.pmap(train_step, "batch", donate_argnums=(0,))
+
+    # Replicate the train state on each device
+    state = jax_utils.replicate(state)
+    unet_params = jax_utils.replicate(unet_params)
+    text_encoder_params = jax_utils.replicate(text_encoder.params)
+    vae_params = jax_utils.replicate(vae_params)
+
+    # Train!
+    if args.streaming:
+        dataset_length = args.max_train_samples
+    else:
+        dataset_length = len(train_dataloader)
+    num_update_steps_per_epoch = math.ceil(dataset_length / args.gradient_accumulation_steps)
+
+    # Scheduler and math around the number of training steps.
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {args.max_train_samples if args.streaming else len(train_dataset)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel & distributed) = {total_train_batch_size}")
+    logger.info(f"  Total optimization steps = {args.num_train_epochs * num_update_steps_per_epoch}")
+
+    if jax.process_index() == 0 and args.report_to == "wandb":
+        wandb.define_metric("*", step_metric="train/step")
+        wandb.define_metric("train/step", step_metric="walltime")
+        wandb.config.update(
+            {
+                "num_train_examples": args.max_train_samples if args.streaming else len(train_dataset),
+                "total_train_batch_size": total_train_batch_size,
+                "total_optimization_step": args.num_train_epochs * num_update_steps_per_epoch,
+                "num_devices": jax.device_count(),
+                "controlnet_params": sum(np.prod(x.shape) for x in jax.tree_util.tree_leaves(state.params)),
+            }
+        )
+
+    global_step = step0 = 0
+    epochs = tqdm(
+        range(args.num_train_epochs),
+        desc="Epoch ... ",
+        position=0,
+        disable=jax.process_index() > 0,
+    )
+    if args.profile_memory:
+        jax.profiler.save_device_memory_profile(os.path.join(args.output_dir, "memory_initial.prof"))
+    t00 = t0 = time.monotonic()
+    for epoch in epochs:
+        # ======================== Training ================================
+
+        train_metrics = []
+        train_metric = None
+
+        steps_per_epoch = (
+            args.max_train_samples // total_train_batch_size
+            if args.streaming or args.max_train_samples
+            else len(train_dataset) // total_train_batch_size
+        )
+        train_step_progress_bar = tqdm(
+            total=steps_per_epoch,
+            desc="Training...",
+            position=1,
+            leave=False,
+            disable=jax.process_index() > 0,
+        )
+        # train
+        for batch in train_dataloader:
+            if args.profile_steps and global_step == 1:
+                train_metric["loss"].block_until_ready()
+                jax.profiler.start_trace(args.output_dir)
+            if args.profile_steps and global_step == 1 + args.profile_steps:
+                train_metric["loss"].block_until_ready()
+                jax.profiler.stop_trace()
+
+            batch = shard(batch)
+            with jax.profiler.StepTraceAnnotation("train", step_num=global_step):
+                state, train_metric, train_rngs = p_train_step(
+                    state, unet_params, text_encoder_params, vae_params, batch, train_rngs
+                )
+            train_metrics.append(train_metric)
+
+            train_step_progress_bar.update(1)
+
+            global_step += 1
+            if global_step >= args.max_train_steps:
+                break
+
+            if (
+                args.validation_prompt is not None
+                and global_step % args.validation_steps == 0
+                and jax.process_index() == 0
+            ):
+                _ = log_validation(
+                    pipeline, pipeline_params, state.params, tokenizer, args, validation_rng, weight_dtype
+                )
+
+            if global_step % args.logging_steps == 0 and jax.process_index() == 0:
+                if args.report_to == "wandb":
+                    train_metrics = jax_utils.unreplicate(train_metrics)
+                    train_metrics = jax.tree_util.tree_map(lambda *m: jnp.array(m).mean(), *train_metrics)
+                    wandb.log(
+                        {
+                            "walltime": time.monotonic() - t00,
+                            "train/step": global_step,
+                            "train/epoch": global_step / dataset_length,
+                            "train/steps_per_sec": (global_step - step0) / (time.monotonic() - t0),
+                            **{f"train/{k}": v for k, v in train_metrics.items()},
+                        }
+                    )
+                t0, step0 = time.monotonic(), global_step
+                train_metrics = []
+            if global_step % args.checkpointing_steps == 0 and jax.process_index() == 0:
+                controlnet.save_pretrained(
+                    f"{args.output_dir}/{global_step}",
+                    params=get_params_to_save(state.params),
+                )
+
+        train_metric = jax_utils.unreplicate(train_metric)
+        train_step_progress_bar.close()
+        epochs.write(f"Epoch... ({epoch + 1}/{args.num_train_epochs} | Loss: {train_metric['loss']})")
+
+    # Final validation & store model.
+    if jax.process_index() == 0:
+        if args.validation_prompt is not None:
+            if args.profile_validation:
+                jax.profiler.start_trace(args.output_dir)
+            image_logs = log_validation(
+                pipeline, pipeline_params, state.params, tokenizer, args, validation_rng, weight_dtype
+            )
+            if args.profile_validation:
+                jax.profiler.stop_trace()
+        else:
+            image_logs = None
+
+        controlnet.save_pretrained(
+            args.output_dir,
+            params=get_params_to_save(state.params),
+        )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    if args.profile_memory:
+        jax.profiler.save_device_memory_profile(os.path.join(args.output_dir, "memory_final.prof"))
+    logger.info("Finished training.")
+
+
+if __name__ == "__main__":
+    main()

train_controlnet_flux.py

@@ -0,0 +1,1435 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+import argparse
+import copy
+import functools
+import logging
+import math
+import os
+import random
+import shutil
+from contextlib import nullcontext
+from pathlib import Path
+
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import DistributedType, ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import (
+    AutoTokenizer,
+    CLIPTextModel,
+    T5EncoderModel,
+)
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    FlowMatchEulerDiscreteScheduler,
+    FluxTransformer2DModel,
+)
+from diffusers.models.controlnet_flux import FluxControlNetModel
+from diffusers.optimization import get_scheduler
+from diffusers.pipelines.flux.pipeline_flux_controlnet import FluxControlNetPipeline
+from diffusers.training_utils import compute_density_for_timestep_sampling, free_memory
+from diffusers.utils import check_min_version, is_wandb_available, make_image_grid
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+from diffusers.utils.import_utils import is_torch_npu_available, is_xformers_available
+from diffusers.utils.torch_utils import is_compiled_module
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.33.0.dev0")
+
+logger = get_logger(__name__)
+if is_torch_npu_available():
+    torch.npu.config.allow_internal_format = False
+
+
+def log_validation(
+    vae, flux_transformer, flux_controlnet, args, accelerator, weight_dtype, step, is_final_validation=False
+):
+    logger.info("Running validation... ")
+
+    if not is_final_validation:
+        flux_controlnet = accelerator.unwrap_model(flux_controlnet)
+        pipeline = FluxControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            controlnet=flux_controlnet,
+            transformer=flux_transformer,
+            torch_dtype=torch.bfloat16,
+        )
+    else:
+        flux_controlnet = FluxControlNetModel.from_pretrained(
+            args.output_dir, torch_dtype=torch.bfloat16, variant=args.save_weight_dtype
+        )
+        pipeline = FluxControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            controlnet=flux_controlnet,
+            transformer=flux_transformer,
+            torch_dtype=torch.bfloat16,
+        )
+
+    pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+    if is_final_validation or torch.backends.mps.is_available():
+        autocast_ctx = nullcontext()
+    else:
+        autocast_ctx = torch.autocast(accelerator.device.type)
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        from diffusers.utils import load_image
+
+        validation_image = load_image(validation_image)
+        # maybe need to inference on 1024 to get a good image
+        validation_image = validation_image.resize((args.resolution, args.resolution))
+
+        images = []
+
+        # pre calculate  prompt embeds, pooled prompt embeds, text ids because t5 does not support autocast
+        prompt_embeds, pooled_prompt_embeds, text_ids = pipeline.encode_prompt(
+            validation_prompt, prompt_2=validation_prompt
+        )
+        for _ in range(args.num_validation_images):
+            with autocast_ctx:
+                # need to fix in pipeline_flux_controlnet
+                image = pipeline(
+                    prompt_embeds=prompt_embeds,
+                    pooled_prompt_embeds=pooled_prompt_embeds,
+                    control_image=validation_image,
+                    num_inference_steps=28,
+                    controlnet_conditioning_scale=0.7,
+                    guidance_scale=3.5,
+                    generator=generator,
+                ).images[0]
+            image = image.resize((args.resolution, args.resolution))
+            images.append(image)
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    tracker_key = "test" if is_final_validation else "validation"
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = [np.asarray(validation_image)]
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({tracker_key: formatted_images})
+        else:
+            logger.warning(f"image logging not implemented for {tracker.name}")
+
+        del pipeline
+        free_memory()
+        return image_logs
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            make_image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+
+## License
+
+Please adhere to the licensing terms as described [here](https://huggingface.co/black-forest-labs/FLUX.1-dev/blob/main/LICENSE.md)
+"""
+
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="other",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "flux",
+        "flux-diffusers",
+        "text-to-image",
+        "diffusers",
+        "controlnet",
+        "diffusers-training",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--pretrained_vae_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to an improved VAE to stabilize training. For more details check out: https://github.com/huggingface/diffusers/pull/4038.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--variant",
+        type=str,
+        default=None,
+        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_h",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_w",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--use_adafactor",
+        action="store_true",
+        help=(
+            "Adafactor is a stochastic optimization method based on Adam that reduces memory usage while retaining"
+            "the empirical benefits of adaptivity. This is achieved through maintaining a factored representation "
+            "of the squared gradient accumulator across training steps."
+        ),
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--enable_npu_flash_attention", action="store_true", help="Whether or not to use npu flash attention."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_double_layers",
+        type=int,
+        default=4,
+        help="Number of double layers in the controlnet (default: 4).",
+    )
+    parser.add_argument(
+        "--num_single_layers",
+        type=int,
+        default=4,
+        help="Number of single layers in the controlnet (default: 4).",
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=2,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="flux_train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+    parser.add_argument(
+        "--jsonl_for_train",
+        type=str,
+        default=None,
+        help="Path to the jsonl file containing the training data.",
+    )
+
+    parser.add_argument(
+        "--guidance_scale",
+        type=float,
+        default=3.5,
+        help="the guidance scale used for transformer.",
+    )
+
+    parser.add_argument(
+        "--save_weight_dtype",
+        type=str,
+        default="fp32",
+        choices=[
+            "fp16",
+            "bf16",
+            "fp32",
+        ],
+        help=("Preserve precision type according to selected weight"),
+    )
+
+    parser.add_argument(
+        "--weighting_scheme",
+        type=str,
+        default="logit_normal",
+        choices=["sigma_sqrt", "logit_normal", "mode", "cosmap", "none"],
+        help=('We default to the "none" weighting scheme for uniform sampling and uniform loss'),
+    )
+    parser.add_argument(
+        "--logit_mean", type=float, default=0.0, help="mean to use when using the `'logit_normal'` weighting scheme."
+    )
+    parser.add_argument(
+        "--logit_std", type=float, default=1.0, help="std to use when using the `'logit_normal'` weighting scheme."
+    )
+    parser.add_argument(
+        "--mode_scale",
+        type=float,
+        default=1.29,
+        help="Scale of mode weighting scheme. Only effective when using the `'mode'` as the `weighting_scheme`.",
+    )
+    parser.add_argument(
+        "--enable_model_cpu_offload",
+        action="store_true",
+        help="Enable model cpu offload and save memory.",
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.jsonl_for_train is None:
+        raise ValueError("Specify either `--dataset_name` or `--jsonl_for_train`")
+
+    if args.dataset_name is not None and args.jsonl_for_train is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--jsonl_for_train`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def get_train_dataset(args, accelerator):
+    dataset = None
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+        )
+    if args.jsonl_for_train is not None:
+        # load from json
+        dataset = load_dataset("json", data_files=args.jsonl_for_train, cache_dir=args.cache_dir)
+        dataset = dataset.flatten_indices()
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    with accelerator.main_process_first():
+        train_dataset = dataset["train"].shuffle(seed=args.seed)
+        if args.max_train_samples is not None:
+            train_dataset = train_dataset.select(range(args.max_train_samples))
+    return train_dataset
+
+
+def prepare_train_dataset(dataset, accelerator):
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [
+            (image.convert("RGB") if not isinstance(image, str) else Image.open(image).convert("RGB"))
+            for image in examples[args.image_column]
+        ]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [
+            (image.convert("RGB") if not isinstance(image, str) else Image.open(image).convert("RGB"))
+            for image in examples[args.conditioning_image_column]
+        ]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+
+        return examples
+
+    with accelerator.main_process_first():
+        dataset = dataset.with_transform(preprocess_train)
+
+    return dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    prompt_ids = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
+
+    pooled_prompt_embeds = torch.stack([torch.tensor(example["pooled_prompt_embeds"]) for example in examples])
+    text_ids = torch.stack([torch.tensor(example["text_ids"]) for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "prompt_ids": prompt_ids,
+        "unet_added_conditions": {"pooled_prompt_embeds": pooled_prompt_embeds, "time_ids": text_ids},
+    }
+
+
+def main(args):
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    logging_out_dir = Path(args.output_dir, args.logging_dir)
+
+    if torch.backends.mps.is_available() and args.mixed_precision == "bf16":
+        # due to pytorch#99272, MPS does not yet support bfloat16.
+        raise ValueError(
+            "Mixed precision training with bfloat16 is not supported on MPS. Please use fp16 (recommended) or fp32 instead."
+        )
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=str(logging_out_dir))
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Disable AMP for MPS. A technique for accelerating machine learning computations on iOS and macOS devices.
+    if torch.backends.mps.is_available():
+        print("MPS is enabled. Disabling AMP.")
+        accelerator.native_amp = False
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        # DEBUG, INFO, WARNING, ERROR, CRITICAL
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizers
+    # load clip tokenizer
+    tokenizer_one = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer",
+        revision=args.revision,
+    )
+    # load t5 tokenizer
+    tokenizer_two = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer_2",
+        revision=args.revision,
+    )
+    # load clip text encoder
+    text_encoder_one = CLIPTextModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    # load t5 text encoder
+    text_encoder_two = T5EncoderModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant
+    )
+
+    vae = AutoencoderKL.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="vae",
+        revision=args.revision,
+        variant=args.variant,
+    )
+    flux_transformer = FluxTransformer2DModel.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="transformer",
+        revision=args.revision,
+        variant=args.variant,
+    )
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        flux_controlnet = FluxControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from transformer")
+        # we can define the num_layers, num_single_layers,
+        flux_controlnet = FluxControlNetModel.from_transformer(
+            flux_transformer,
+            attention_head_dim=flux_transformer.config["attention_head_dim"],
+            num_attention_heads=flux_transformer.config["num_attention_heads"],
+            num_layers=args.num_double_layers,
+            num_single_layers=args.num_single_layers,
+        )
+    logger.info("all models loaded successfully")
+
+    noise_scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="scheduler",
+    )
+    noise_scheduler_copy = copy.deepcopy(noise_scheduler)
+    vae.requires_grad_(False)
+    flux_transformer.requires_grad_(False)
+    text_encoder_one.requires_grad_(False)
+    text_encoder_two.requires_grad_(False)
+    flux_controlnet.train()
+
+    # use some pipeline function
+    flux_controlnet_pipeline = FluxControlNetPipeline(
+        scheduler=noise_scheduler,
+        vae=vae,
+        text_encoder=text_encoder_one,
+        tokenizer=tokenizer_one,
+        text_encoder_2=text_encoder_two,
+        tokenizer_2=tokenizer_two,
+        transformer=flux_transformer,
+        controlnet=flux_controlnet,
+    )
+    if args.enable_model_cpu_offload:
+        flux_controlnet_pipeline.enable_model_cpu_offload()
+    else:
+        flux_controlnet_pipeline.to(accelerator.device)
+
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            if accelerator.is_main_process:
+                i = len(weights) - 1
+
+                while len(weights) > 0:
+                    weights.pop()
+                    model = models[i]
+
+                    sub_dir = "flux_controlnet"
+                    model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                    i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = FluxControlNetModel.from_pretrained(input_dir, subfolder="flux_controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    if args.enable_npu_flash_attention:
+        if is_torch_npu_available():
+            logger.info("npu flash attention enabled.")
+            flux_transformer.enable_npu_flash_attention()
+        else:
+            raise ValueError("npu flash attention requires torch_npu extensions and is supported only on npu devices.")
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warning(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            flux_transformer.enable_xformers_memory_efficient_attention()
+            flux_controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        flux_transformer.enable_gradient_checkpointing()
+        flux_controlnet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if unwrap_model(flux_controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {unwrap_model(flux_controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = flux_controlnet.parameters()
+    # use adafactor optimizer to save gpu memory
+    if args.use_adafactor:
+        from transformers import Adafactor
+
+        optimizer = Adafactor(
+            params_to_optimize,
+            lr=args.learning_rate,
+            scale_parameter=False,
+            relative_step=False,
+            # warmup_init=True,
+            weight_decay=args.adam_weight_decay,
+        )
+    else:
+        optimizer = optimizer_class(
+            params_to_optimize,
+            lr=args.learning_rate,
+            betas=(args.adam_beta1, args.adam_beta2),
+            weight_decay=args.adam_weight_decay,
+            eps=args.adam_epsilon,
+        )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    vae.to(accelerator.device, dtype=weight_dtype)
+    flux_transformer.to(accelerator.device, dtype=weight_dtype)
+
+    def compute_embeddings(batch, proportion_empty_prompts, flux_controlnet_pipeline, weight_dtype, is_train=True):
+        prompt_batch = batch[args.caption_column]
+        captions = []
+        for caption in prompt_batch:
+            if random.random() < proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+        prompt_batch = captions
+        prompt_embeds, pooled_prompt_embeds, text_ids = flux_controlnet_pipeline.encode_prompt(
+            prompt_batch, prompt_2=prompt_batch
+        )
+        prompt_embeds = prompt_embeds.to(dtype=weight_dtype)
+        pooled_prompt_embeds = pooled_prompt_embeds.to(dtype=weight_dtype)
+        text_ids = text_ids.to(dtype=weight_dtype)
+
+        # text_ids [512,3] to [bs,512,3]
+        text_ids = text_ids.unsqueeze(0).expand(prompt_embeds.shape[0], -1, -1)
+        return {"prompt_embeds": prompt_embeds, "pooled_prompt_embeds": pooled_prompt_embeds, "text_ids": text_ids}
+
+    train_dataset = get_train_dataset(args, accelerator)
+    text_encoders = [text_encoder_one, text_encoder_two]
+    tokenizers = [tokenizer_one, tokenizer_two]
+    compute_embeddings_fn = functools.partial(
+        compute_embeddings,
+        flux_controlnet_pipeline=flux_controlnet_pipeline,
+        proportion_empty_prompts=args.proportion_empty_prompts,
+        weight_dtype=weight_dtype,
+    )
+    with accelerator.main_process_first():
+        from datasets.fingerprint import Hasher
+
+        # fingerprint used by the cache for the other processes to load the result
+        # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
+        new_fingerprint = Hasher.hash(args)
+        train_dataset = train_dataset.map(
+            compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint, batch_size=50
+        )
+
+    del text_encoders, tokenizers, text_encoder_one, text_encoder_two, tokenizer_one, tokenizer_two
+    free_memory()
+
+    # Then get the training dataset ready to be passed to the dataloader.
+    train_dataset = prepare_train_dataset(train_dataset, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    # Check the PR https://github.com/huggingface/diffusers/pull/8312 for detailed explanation.
+    if args.max_train_steps is None:
+        len_train_dataloader_after_sharding = math.ceil(len(train_dataloader) / accelerator.num_processes)
+        num_update_steps_per_epoch = math.ceil(len_train_dataloader_after_sharding / args.gradient_accumulation_steps)
+        num_training_steps_for_scheduler = (
+            args.num_train_epochs * num_update_steps_per_epoch * accelerator.num_processes
+        )
+    else:
+        num_training_steps_for_scheduler = args.max_train_steps * accelerator.num_processes
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
+        num_training_steps=args.max_train_steps * accelerator.num_processes,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+    # Prepare everything with our `accelerator`.
+    flux_controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        flux_controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        if num_training_steps_for_scheduler != args.max_train_steps * accelerator.num_processes:
+            logger.warning(
+                f"The length of the 'train_dataloader' after 'accelerator.prepare' ({len(train_dataloader)}) does not match "
+                f"the expected length ({len_train_dataloader_after_sharding}) when the learning rate scheduler was created. "
+                f"This inconsistency may result in the learning rate scheduler not functioning properly."
+            )
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
+        sigmas = noise_scheduler_copy.sigmas.to(device=accelerator.device, dtype=dtype)
+        schedule_timesteps = noise_scheduler_copy.timesteps.to(accelerator.device)
+        timesteps = timesteps.to(accelerator.device)
+        step_indices = [(schedule_timesteps == t).nonzero().item() for t in timesteps]
+
+        sigma = sigmas[step_indices].flatten()
+        while len(sigma.shape) < n_dim:
+            sigma = sigma.unsqueeze(-1)
+        return sigma
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(flux_controlnet):
+                # Convert images to latent space
+                # vae encode
+                pixel_values = batch["pixel_values"].to(dtype=weight_dtype)
+                pixel_latents_tmp = vae.encode(pixel_values).latent_dist.sample()
+                pixel_latents_tmp = (pixel_latents_tmp - vae.config.shift_factor) * vae.config.scaling_factor
+                pixel_latents = FluxControlNetPipeline._pack_latents(
+                    pixel_latents_tmp,
+                    pixel_values.shape[0],
+                    pixel_latents_tmp.shape[1],
+                    pixel_latents_tmp.shape[2],
+                    pixel_latents_tmp.shape[3],
+                )
+
+                control_values = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+                control_latents = vae.encode(control_values).latent_dist.sample()
+                control_latents = (control_latents - vae.config.shift_factor) * vae.config.scaling_factor
+                control_image = FluxControlNetPipeline._pack_latents(
+                    control_latents,
+                    control_values.shape[0],
+                    control_latents.shape[1],
+                    control_latents.shape[2],
+                    control_latents.shape[3],
+                )
+
+                latent_image_ids = FluxControlNetPipeline._prepare_latent_image_ids(
+                    batch_size=pixel_latents_tmp.shape[0],
+                    height=pixel_latents_tmp.shape[2] // 2,
+                    width=pixel_latents_tmp.shape[3] // 2,
+                    device=pixel_values.device,
+                    dtype=pixel_values.dtype,
+                )
+
+                bsz = pixel_latents.shape[0]
+                noise = torch.randn_like(pixel_latents).to(accelerator.device).to(dtype=weight_dtype)
+                # Sample a random timestep for each image
+                # for weighting schemes where we sample timesteps non-uniformly
+                u = compute_density_for_timestep_sampling(
+                    weighting_scheme=args.weighting_scheme,
+                    batch_size=bsz,
+                    logit_mean=args.logit_mean,
+                    logit_std=args.logit_std,
+                    mode_scale=args.mode_scale,
+                )
+                indices = (u * noise_scheduler_copy.config.num_train_timesteps).long()
+                timesteps = noise_scheduler_copy.timesteps[indices].to(device=pixel_latents.device)
+
+                # Add noise according to flow matching.
+                sigmas = get_sigmas(timesteps, n_dim=pixel_latents.ndim, dtype=pixel_latents.dtype)
+                noisy_model_input = (1.0 - sigmas) * pixel_latents + sigmas * noise
+
+                # handle guidance
+                if flux_transformer.config.guidance_embeds:
+                    guidance_vec = torch.full(
+                        (noisy_model_input.shape[0],),
+                        args.guidance_scale,
+                        device=noisy_model_input.device,
+                        dtype=weight_dtype,
+                    )
+                else:
+                    guidance_vec = None
+
+                controlnet_block_samples, controlnet_single_block_samples = flux_controlnet(
+                    hidden_states=noisy_model_input,
+                    controlnet_cond=control_image,
+                    timestep=timesteps / 1000,
+                    guidance=guidance_vec,
+                    pooled_projections=batch["unet_added_conditions"]["pooled_prompt_embeds"].to(dtype=weight_dtype),
+                    encoder_hidden_states=batch["prompt_ids"].to(dtype=weight_dtype),
+                    txt_ids=batch["unet_added_conditions"]["time_ids"][0].to(dtype=weight_dtype),
+                    img_ids=latent_image_ids,
+                    return_dict=False,
+                )
+
+                noise_pred = flux_transformer(
+                    hidden_states=noisy_model_input,
+                    timestep=timesteps / 1000,
+                    guidance=guidance_vec,
+                    pooled_projections=batch["unet_added_conditions"]["pooled_prompt_embeds"].to(dtype=weight_dtype),
+                    encoder_hidden_states=batch["prompt_ids"].to(dtype=weight_dtype),
+                    controlnet_block_samples=[sample.to(dtype=weight_dtype) for sample in controlnet_block_samples]
+                    if controlnet_block_samples is not None
+                    else None,
+                    controlnet_single_block_samples=[
+                        sample.to(dtype=weight_dtype) for sample in controlnet_single_block_samples
+                    ]
+                    if controlnet_single_block_samples is not None
+                    else None,
+                    txt_ids=batch["unet_added_conditions"]["time_ids"][0].to(dtype=weight_dtype),
+                    img_ids=latent_image_ids,
+                    return_dict=False,
+                )[0]
+
+                loss = F.mse_loss(noise_pred.float(), (noise - pixel_latents).float(), reduction="mean")
+                accelerator.backward(loss)
+                # Check if the gradient of each model parameter contains NaN
+                for name, param in flux_controlnet.named_parameters():
+                    if param.grad is not None and torch.isnan(param.grad).any():
+                        logger.error(f"Gradient for {name} contains NaN!")
+
+                if accelerator.sync_gradients:
+                    params_to_clip = flux_controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                # DeepSpeed requires saving weights on every device; saving weights only on the main process would cause issues.
+                if accelerator.distributed_type == DistributedType.DEEPSPEED or accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae=vae,
+                            flux_transformer=flux_transformer,
+                            flux_controlnet=flux_controlnet,
+                            args=args,
+                            accelerator=accelerator,
+                            weight_dtype=weight_dtype,
+                            step=global_step,
+                        )
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        flux_controlnet = unwrap_model(flux_controlnet)
+        save_weight_dtype = torch.float32
+        if args.save_weight_dtype == "fp16":
+            save_weight_dtype = torch.float16
+        elif args.save_weight_dtype == "bf16":
+            save_weight_dtype = torch.bfloat16
+        flux_controlnet.to(save_weight_dtype)
+        if args.save_weight_dtype != "fp32":
+            flux_controlnet.save_pretrained(args.output_dir, variant=args.save_weight_dtype)
+        else:
+            flux_controlnet.save_pretrained(args.output_dir)
+        # Run a final round of validation.
+        # Setting `vae`, `unet`, and `controlnet` to None to load automatically from `args.output_dir`.
+        image_logs = None
+        if args.validation_prompt is not None:
+            image_logs = log_validation(
+                vae=vae,
+                flux_transformer=flux_transformer,
+                flux_controlnet=None,
+                args=args,
+                accelerator=accelerator,
+                weight_dtype=weight_dtype,
+                step=global_step,
+                is_final_validation=True,
+            )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

train_controlnet_sd3.py

@@ -0,0 +1,1428 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+import argparse
+import contextlib
+import copy
+import functools
+import logging
+import math
+import os
+import random
+import shutil
+from pathlib import Path
+
+import accelerate
+import numpy as np
+import torch
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import DistributedDataParallelKwargs, ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import CLIPTokenizer, PretrainedConfig, T5TokenizerFast
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    FlowMatchEulerDiscreteScheduler,
+    SD3ControlNetModel,
+    SD3Transformer2DModel,
+    StableDiffusion3ControlNetPipeline,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.training_utils import compute_density_for_timestep_sampling, compute_loss_weighting_for_sd3, free_memory
+from diffusers.utils import check_min_version, is_wandb_available, make_image_grid
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+from diffusers.utils.torch_utils import is_compiled_module
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.33.0.dev0")
+
+logger = get_logger(__name__)
+
+
+def log_validation(controlnet, args, accelerator, weight_dtype, step, is_final_validation=False):
+    logger.info("Running validation... ")
+
+    if not is_final_validation:
+        controlnet = accelerator.unwrap_model(controlnet)
+    else:
+        controlnet = SD3ControlNetModel.from_pretrained(args.output_dir, torch_dtype=weight_dtype)
+
+    pipeline = StableDiffusion3ControlNetPipeline.from_pretrained(
+        args.pretrained_model_name_or_path,
+        controlnet=controlnet,
+        safety_checker=None,
+        revision=args.revision,
+        variant=args.variant,
+        torch_dtype=weight_dtype,
+    )
+    pipeline = pipeline.to(torch.device(accelerator.device))
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+    inference_ctx = contextlib.nullcontext() if is_final_validation else torch.autocast(accelerator.device.type)
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        validation_image = Image.open(validation_image).convert("RGB")
+
+        images = []
+
+        for _ in range(args.num_validation_images):
+            with inference_ctx:
+                image = pipeline(
+                    validation_prompt, control_image=validation_image, num_inference_steps=20, generator=generator
+                ).images[0]
+
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    tracker_key = "test" if is_final_validation else "validation"
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                tracker.writer.add_image(
+                    "Controlnet conditioning", np.asarray([validation_image]), step, dataformats="NHWC"
+                )
+
+                formatted_images = []
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({tracker_key: formatted_images})
+        else:
+            logger.warning(f"image logging not implemented for {tracker.name}")
+
+        del pipeline
+        free_memory()
+
+        if not is_final_validation:
+            controlnet.to(accelerator.device)
+
+        return image_logs
+
+
+# Copied from dreambooth sd3 example
+def load_text_encoders(class_one, class_two, class_three):
+    text_encoder_one = class_one.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    text_encoder_two = class_two.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant
+    )
+    text_encoder_three = class_three.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_3", revision=args.revision, variant=args.variant
+    )
+    return text_encoder_one, text_encoder_two, text_encoder_three
+
+
+# Copied from dreambooth sd3 example
+def import_model_class_from_model_name_or_path(
+    pretrained_model_name_or_path: str, revision: str, subfolder: str = "text_encoder"
+):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path, subfolder=subfolder, revision=revision
+    )
+    model_class = text_encoder_config.architectures[0]
+    if model_class == "CLIPTextModelWithProjection":
+        from transformers import CLIPTextModelWithProjection
+
+        return CLIPTextModelWithProjection
+    elif model_class == "T5EncoderModel":
+        from transformers import T5EncoderModel
+
+        return T5EncoderModel
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            make_image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# SD3 controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+The weights were trained using [ControlNet](https://github.com/lllyasviel/ControlNet) with the [SD3 diffusers trainer](https://github.com/huggingface/diffusers/blob/main/examples/controlnet/README_sd3.md).
+{img_str}
+
+Please adhere to the licensing terms as described `[here](https://huggingface.co/stabilityai/stable-diffusion-3-medium/blob/main/LICENSE)`.
+"""
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="openrail++",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "text-to-image",
+        "diffusers-training",
+        "diffusers",
+        "sd3",
+        "sd3-diffusers",
+        "controlnet",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--num_extra_conditioning_channels",
+        type=int,
+        default=0,
+        help="Number of extra conditioning channels for controlnet.",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--variant",
+        type=str,
+        default=None,
+        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--upcast_vae",
+        action="store_true",
+        help="Whether or not to upcast vae to fp32",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument(
+        "--weighting_scheme",
+        type=str,
+        default="logit_normal",
+        choices=["sigma_sqrt", "logit_normal", "mode", "cosmap"],
+    )
+    parser.add_argument(
+        "--logit_mean", type=float, default=0.0, help="mean to use when using the `'logit_normal'` weighting scheme."
+    )
+    parser.add_argument(
+        "--logit_std", type=float, default=1.0, help="std to use when using the `'logit_normal'` weighting scheme."
+    )
+    parser.add_argument(
+        "--mode_scale",
+        type=float,
+        default=1.29,
+        help="Scale of mode weighting scheme. Only effective when using the `'mode'` as the `weighting_scheme`.",
+    )
+    parser.add_argument(
+        "--precondition_outputs",
+        type=int,
+        default=1,
+        help="Flag indicating if we are preconditioning the model outputs or not as done in EDM. This affects how "
+        "model `target` is calculated.",
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--max_sequence_length",
+        type=int,
+        default=77,
+        help="Maximum sequence length to use with with the T5 text encoder",
+    )
+    parser.add_argument(
+        "--dataset_preprocess_batch_size", type=int, default=1000, help="Batch size for preprocessing dataset."
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.dataset_name is not None and args.train_data_dir is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def make_train_dataset(args, tokenizer_one, tokenizer_two, tokenizer_three, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    def process_captions(examples, is_train=True):
+        captions = []
+        for caption in examples[caption_column]:
+            if random.random() < args.proportion_empty_prompts:
+                captions.append("")
+            elif isinstance(caption, str):
+                captions.append(caption)
+            elif isinstance(caption, (list, np.ndarray)):
+                # take a random caption if there are multiple
+                captions.append(random.choice(caption) if is_train else caption[0])
+            else:
+                raise ValueError(
+                    f"Caption column `{caption_column}` should contain either strings or lists of strings."
+                )
+        return captions
+
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [image.convert("RGB") for image in examples[image_column]]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [image.convert("RGB") for image in examples[conditioning_image_column]]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+        examples["prompts"] = process_captions(examples)
+
+        return examples
+
+    with accelerator.main_process_first():
+        if args.max_train_samples is not None:
+            dataset["train"] = dataset["train"].shuffle(seed=args.seed).select(range(args.max_train_samples))
+        # Set the training transforms
+        train_dataset = dataset["train"].with_transform(preprocess_train)
+
+    return train_dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    prompt_embeds = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
+    pooled_prompt_embeds = torch.stack([torch.tensor(example["pooled_prompt_embeds"]) for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "prompt_embeds": prompt_embeds,
+        "pooled_prompt_embeds": pooled_prompt_embeds,
+    }
+
+
+# Copied from dreambooth sd3 example
+def _encode_prompt_with_t5(
+    text_encoder,
+    tokenizer,
+    max_sequence_length,
+    prompt=None,
+    num_images_per_prompt=1,
+    device=None,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+    batch_size = len(prompt)
+
+    text_inputs = tokenizer(
+        prompt,
+        padding="max_length",
+        max_length=max_sequence_length,
+        truncation=True,
+        add_special_tokens=True,
+        return_tensors="pt",
+    )
+    text_input_ids = text_inputs.input_ids
+    prompt_embeds = text_encoder(text_input_ids.to(device))[0]
+
+    dtype = text_encoder.dtype
+    prompt_embeds = prompt_embeds.to(dtype=dtype, device=device)
+
+    _, seq_len, _ = prompt_embeds.shape
+
+    # duplicate text embeddings and attention mask for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+    return prompt_embeds
+
+
+# Copied from dreambooth sd3 example
+def _encode_prompt_with_clip(
+    text_encoder,
+    tokenizer,
+    prompt: str,
+    device=None,
+    num_images_per_prompt: int = 1,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+    batch_size = len(prompt)
+
+    text_inputs = tokenizer(
+        prompt,
+        padding="max_length",
+        max_length=77,
+        truncation=True,
+        return_tensors="pt",
+    )
+
+    text_input_ids = text_inputs.input_ids
+    prompt_embeds = text_encoder(text_input_ids.to(device), output_hidden_states=True)
+
+    pooled_prompt_embeds = prompt_embeds[0]
+    prompt_embeds = prompt_embeds.hidden_states[-2]
+    prompt_embeds = prompt_embeds.to(dtype=text_encoder.dtype, device=device)
+
+    _, seq_len, _ = prompt_embeds.shape
+    # duplicate text embeddings for each generation per prompt, using mps friendly method
+    prompt_embeds = prompt_embeds.repeat(1, num_images_per_prompt, 1)
+    prompt_embeds = prompt_embeds.view(batch_size * num_images_per_prompt, seq_len, -1)
+
+    return prompt_embeds, pooled_prompt_embeds
+
+
+# Copied from dreambooth sd3 example
+def encode_prompt(
+    text_encoders,
+    tokenizers,
+    prompt: str,
+    max_sequence_length,
+    device=None,
+    num_images_per_prompt: int = 1,
+):
+    prompt = [prompt] if isinstance(prompt, str) else prompt
+
+    clip_tokenizers = tokenizers[:2]
+    clip_text_encoders = text_encoders[:2]
+
+    clip_prompt_embeds_list = []
+    clip_pooled_prompt_embeds_list = []
+    for tokenizer, text_encoder in zip(clip_tokenizers, clip_text_encoders):
+        prompt_embeds, pooled_prompt_embeds = _encode_prompt_with_clip(
+            text_encoder=text_encoder,
+            tokenizer=tokenizer,
+            prompt=prompt,
+            device=device if device is not None else text_encoder.device,
+            num_images_per_prompt=num_images_per_prompt,
+        )
+        clip_prompt_embeds_list.append(prompt_embeds)
+        clip_pooled_prompt_embeds_list.append(pooled_prompt_embeds)
+
+    clip_prompt_embeds = torch.cat(clip_prompt_embeds_list, dim=-1)
+    pooled_prompt_embeds = torch.cat(clip_pooled_prompt_embeds_list, dim=-1)
+
+    t5_prompt_embed = _encode_prompt_with_t5(
+        text_encoders[-1],
+        tokenizers[-1],
+        max_sequence_length,
+        prompt=prompt,
+        num_images_per_prompt=num_images_per_prompt,
+        device=device if device is not None else text_encoders[-1].device,
+    )
+
+    clip_prompt_embeds = torch.nn.functional.pad(
+        clip_prompt_embeds, (0, t5_prompt_embed.shape[-1] - clip_prompt_embeds.shape[-1])
+    )
+    prompt_embeds = torch.cat([clip_prompt_embeds, t5_prompt_embed], dim=-2)
+
+    return prompt_embeds, pooled_prompt_embeds
+
+
+def main(args):
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    if torch.backends.mps.is_available() and args.mixed_precision == "bf16":
+        # due to pytorch#99272, MPS does not yet support bfloat16.
+        raise ValueError(
+            "Mixed precision training with bfloat16 is not supported on MPS. Please use fp16 (recommended) or fp32 instead."
+        )
+
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+    kwargs = DistributedDataParallelKwargs(find_unused_parameters=True)
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+        kwargs_handlers=[kwargs],
+    )
+
+    # Disable AMP for MPS.
+    if torch.backends.mps.is_available():
+        accelerator.native_amp = False
+
+    if args.report_to == "wandb":
+        if not is_wandb_available():
+            raise ImportError("Make sure to install wandb if you want to use it for logging during training.")
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizer
+    tokenizer_one = CLIPTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer",
+        revision=args.revision,
+    )
+    tokenizer_two = CLIPTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer_2",
+        revision=args.revision,
+    )
+    tokenizer_three = T5TokenizerFast.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer_3",
+        revision=args.revision,
+    )
+
+    # import correct text encoder class
+    text_encoder_cls_one = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision
+    )
+    text_encoder_cls_two = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision, subfolder="text_encoder_2"
+    )
+    text_encoder_cls_three = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision, subfolder="text_encoder_3"
+    )
+
+    # Load scheduler and models
+    noise_scheduler = FlowMatchEulerDiscreteScheduler.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="scheduler"
+    )
+    noise_scheduler_copy = copy.deepcopy(noise_scheduler)
+    text_encoder_one, text_encoder_two, text_encoder_three = load_text_encoders(
+        text_encoder_cls_one, text_encoder_cls_two, text_encoder_cls_three
+    )
+    vae = AutoencoderKL.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="vae",
+        revision=args.revision,
+        variant=args.variant,
+    )
+    transformer = SD3Transformer2DModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="transformer", revision=args.revision, variant=args.variant
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = SD3ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from transformer")
+        controlnet = SD3ControlNetModel.from_transformer(
+            transformer, num_extra_conditioning_channels=args.num_extra_conditioning_channels
+        )
+
+    transformer.requires_grad_(False)
+    vae.requires_grad_(False)
+    text_encoder_one.requires_grad_(False)
+    text_encoder_two.requires_grad_(False)
+    text_encoder_three.requires_grad_(False)
+    controlnet.train()
+
+    # Taken from [Sayak Paul's Diffusers PR #6511](https://github.com/huggingface/diffusers/pull/6511/files)
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            if accelerator.is_main_process:
+                i = len(weights) - 1
+
+                while len(weights) > 0:
+                    weights.pop()
+                    model = models[i]
+
+                    sub_dir = "controlnet"
+                    model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                    i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = SD3ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, transformer and text_encoder to device and cast to weight_dtype
+    if args.upcast_vae:
+        vae.to(accelerator.device, dtype=torch.float32)
+    else:
+        vae.to(accelerator.device, dtype=weight_dtype)
+    transformer.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_one.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_two.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_three.to(accelerator.device, dtype=weight_dtype)
+
+    train_dataset = make_train_dataset(args, tokenizer_one, tokenizer_two, tokenizer_three, accelerator)
+
+    tokenizers = [tokenizer_one, tokenizer_two, tokenizer_three]
+    text_encoders = [text_encoder_one, text_encoder_two, text_encoder_three]
+
+    def compute_text_embeddings(batch, text_encoders, tokenizers):
+        with torch.no_grad():
+            prompt = batch["prompts"]
+            prompt_embeds, pooled_prompt_embeds = encode_prompt(
+                text_encoders, tokenizers, prompt, args.max_sequence_length
+            )
+            prompt_embeds = prompt_embeds.to(accelerator.device)
+            pooled_prompt_embeds = pooled_prompt_embeds.to(accelerator.device)
+        return {"prompt_embeds": prompt_embeds, "pooled_prompt_embeds": pooled_prompt_embeds}
+
+    compute_embeddings_fn = functools.partial(
+        compute_text_embeddings,
+        text_encoders=text_encoders,
+        tokenizers=tokenizers,
+    )
+    with accelerator.main_process_first():
+        from datasets.fingerprint import Hasher
+
+        # fingerprint used by the cache for the other processes to load the result
+        # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
+        new_fingerprint = Hasher.hash(args)
+        train_dataset = train_dataset.map(
+            compute_embeddings_fn,
+            batched=True,
+            batch_size=args.dataset_preprocess_batch_size,
+            new_fingerprint=new_fingerprint,
+        )
+
+    del text_encoder_one, text_encoder_two, text_encoder_three
+    del tokenizer_one, tokenizer_two, tokenizer_three
+    free_memory()
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    overrode_max_train_steps = False
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        overrode_max_train_steps = True
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=args.lr_warmup_steps * accelerator.num_processes,
+        num_training_steps=args.max_train_steps * accelerator.num_processes,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if overrode_max_train_steps:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    def get_sigmas(timesteps, n_dim=4, dtype=torch.float32):
+        sigmas = noise_scheduler_copy.sigmas.to(device=accelerator.device, dtype=dtype)
+        schedule_timesteps = noise_scheduler_copy.timesteps.to(accelerator.device)
+        timesteps = timesteps.to(accelerator.device)
+        step_indices = [(schedule_timesteps == t).nonzero().item() for t in timesteps]
+
+        sigma = sigmas[step_indices].flatten()
+        while len(sigma.shape) < n_dim:
+            sigma = sigma.unsqueeze(-1)
+        return sigma
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                pixel_values = batch["pixel_values"].to(dtype=vae.dtype)
+                model_input = vae.encode(pixel_values).latent_dist.sample()
+                model_input = (model_input - vae.config.shift_factor) * vae.config.scaling_factor
+                model_input = model_input.to(dtype=weight_dtype)
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(model_input)
+                bsz = model_input.shape[0]
+                # Sample a random timestep for each image
+                # for weighting schemes where we sample timesteps non-uniformly
+                u = compute_density_for_timestep_sampling(
+                    weighting_scheme=args.weighting_scheme,
+                    batch_size=bsz,
+                    logit_mean=args.logit_mean,
+                    logit_std=args.logit_std,
+                    mode_scale=args.mode_scale,
+                )
+                indices = (u * noise_scheduler_copy.config.num_train_timesteps).long()
+                timesteps = noise_scheduler_copy.timesteps[indices].to(device=model_input.device)
+
+                # Add noise according to flow matching.
+                # zt = (1 - texp) * x + texp * z1
+                sigmas = get_sigmas(timesteps, n_dim=model_input.ndim, dtype=model_input.dtype)
+                noisy_model_input = (1.0 - sigmas) * model_input + sigmas * noise
+
+                # Get the text embedding for conditioning
+                prompt_embeds = batch["prompt_embeds"].to(dtype=weight_dtype)
+                pooled_prompt_embeds = batch["pooled_prompt_embeds"].to(dtype=weight_dtype)
+
+                # controlnet(s) inference
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+                controlnet_image = vae.encode(controlnet_image).latent_dist.sample()
+                controlnet_image = controlnet_image * vae.config.scaling_factor
+
+                control_block_res_samples = controlnet(
+                    hidden_states=noisy_model_input,
+                    timestep=timesteps,
+                    encoder_hidden_states=prompt_embeds,
+                    pooled_projections=pooled_prompt_embeds,
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )[0]
+                control_block_res_samples = [sample.to(dtype=weight_dtype) for sample in control_block_res_samples]
+
+                # Predict the noise residual
+                model_pred = transformer(
+                    hidden_states=noisy_model_input,
+                    timestep=timesteps,
+                    encoder_hidden_states=prompt_embeds,
+                    pooled_projections=pooled_prompt_embeds,
+                    block_controlnet_hidden_states=control_block_res_samples,
+                    return_dict=False,
+                )[0]
+
+                # Follow: Section 5 of https://arxiv.org/abs/2206.00364.
+                # Preconditioning of the model outputs.
+                if args.precondition_outputs:
+                    model_pred = model_pred * (-sigmas) + noisy_model_input
+
+                # these weighting schemes use a uniform timestep sampling
+                # and instead post-weight the loss
+                weighting = compute_loss_weighting_for_sd3(weighting_scheme=args.weighting_scheme, sigmas=sigmas)
+
+                # flow matching loss
+                if args.precondition_outputs:
+                    target = model_input
+                else:
+                    target = noise - model_input
+
+                # Compute regular loss.
+                loss = torch.mean(
+                    (weighting.float() * (model_pred.float() - target.float()) ** 2).reshape(target.shape[0], -1),
+                    1,
+                )
+                loss = loss.mean()
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                if accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            controlnet,
+                            args,
+                            accelerator,
+                            weight_dtype,
+                            global_step,
+                        )
+
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        # Run a final round of validation.
+        image_logs = None
+        if args.validation_prompt is not None:
+            image_logs = log_validation(
+                controlnet=None,
+                args=args,
+                accelerator=accelerator,
+                weight_dtype=weight_dtype,
+                step=global_step,
+                is_final_validation=True,
+            )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)

train_controlnet_sdxl.py

@@ -0,0 +1,1344 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2025 The HuggingFace Inc. team. All rights reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+
+import argparse
+import functools
+import gc
+import logging
+import math
+import os
+import random
+import shutil
+from contextlib import nullcontext
+from pathlib import Path
+
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import DistributedType, ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, PretrainedConfig
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    ControlNetModel,
+    DDPMScheduler,
+    StableDiffusionXLControlNetPipeline,
+    UNet2DConditionModel,
+    UniPCMultistepScheduler,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available, make_image_grid
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+from diffusers.utils.import_utils import is_torch_npu_available, is_xformers_available
+from diffusers.utils.torch_utils import is_compiled_module
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.33.0.dev0")
+
+logger = get_logger(__name__)
+if is_torch_npu_available():
+    torch.npu.config.allow_internal_format = False
+
+
+def log_validation(vae, unet, controlnet, args, accelerator, weight_dtype, step, is_final_validation=False):
+    logger.info("Running validation... ")
+
+    if not is_final_validation:
+        controlnet = accelerator.unwrap_model(controlnet)
+        pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            vae=vae,
+            unet=unet,
+            controlnet=controlnet,
+            revision=args.revision,
+            variant=args.variant,
+            torch_dtype=weight_dtype,
+        )
+    else:
+        controlnet = ControlNetModel.from_pretrained(args.output_dir, torch_dtype=weight_dtype)
+        if args.pretrained_vae_model_name_or_path is not None:
+            vae = AutoencoderKL.from_pretrained(args.pretrained_vae_model_name_or_path, torch_dtype=weight_dtype)
+        else:
+            vae = AutoencoderKL.from_pretrained(
+                args.pretrained_model_name_or_path, subfolder="vae", torch_dtype=weight_dtype
+            )
+
+        pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            vae=vae,
+            controlnet=controlnet,
+            revision=args.revision,
+            variant=args.variant,
+            torch_dtype=weight_dtype,
+        )
+
+    pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+    if is_final_validation or torch.backends.mps.is_available():
+        autocast_ctx = nullcontext()
+    else:
+        autocast_ctx = torch.autocast(accelerator.device.type)
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        validation_image = Image.open(validation_image).convert("RGB")
+        validation_image = validation_image.resize((args.resolution, args.resolution))
+
+        images = []
+
+        for _ in range(args.num_validation_images):
+            with autocast_ctx:
+                image = pipeline(
+                    prompt=validation_prompt, image=validation_image, num_inference_steps=20, generator=generator
+                ).images[0]
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    tracker_key = "test" if is_final_validation else "validation"
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = [np.asarray(validation_image)]
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({tracker_key: formatted_images})
+        else:
+            logger.warning(f"image logging not implemented for {tracker.name}")
+
+        del pipeline
+        gc.collect()
+        torch.cuda.empty_cache()
+
+        return image_logs
+
+
+def import_model_class_from_model_name_or_path(
+    pretrained_model_name_or_path: str, revision: str, subfolder: str = "text_encoder"
+):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path, subfolder=subfolder, revision=revision
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "CLIPTextModelWithProjection":
+        from transformers import CLIPTextModelWithProjection
+
+        return CLIPTextModelWithProjection
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            make_image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+"""
+
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="openrail++",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "stable-diffusion-xl",
+        "stable-diffusion-xl-diffusers",
+        "text-to-image",
+        "diffusers",
+        "controlnet",
+        "diffusers-training",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--pretrained_vae_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to an improved VAE to stabilize training. For more details check out: https://github.com/huggingface/diffusers/pull/4038.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--variant",
+        type=str,
+        default=None,
+        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_h",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_w",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--enable_npu_flash_attention", action="store_true", help="Whether or not to use npu flash attention."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="sd_xl_train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def get_train_dataset(args, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+            data_dir=args.train_data_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    with accelerator.main_process_first():
+        train_dataset = dataset["train"].shuffle(seed=args.seed)
+        if args.max_train_samples is not None:
+            train_dataset = train_dataset.select(range(args.max_train_samples))
+    return train_dataset
+
+
+# Adapted from pipelines.StableDiffusionXLPipeline.encode_prompt
+def encode_prompt(prompt_batch, text_encoders, tokenizers, proportion_empty_prompts, is_train=True):
+    prompt_embeds_list = []
+
+    captions = []
+    for caption in prompt_batch:
+        if random.random() < proportion_empty_prompts:
+            captions.append("")
+        elif isinstance(caption, str):
+            captions.append(caption)
+        elif isinstance(caption, (list, np.ndarray)):
+            # take a random caption if there are multiple
+            captions.append(random.choice(caption) if is_train else caption[0])
+
+    with torch.no_grad():
+        for tokenizer, text_encoder in zip(tokenizers, text_encoders):
+            text_inputs = tokenizer(
+                captions,
+                padding="max_length",
+                max_length=tokenizer.model_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids
+            prompt_embeds = text_encoder(
+                text_input_ids.to(text_encoder.device),
+                output_hidden_states=True,
+            )
+
+            # We are only ALWAYS interested in the pooled output of the final text encoder
+            pooled_prompt_embeds = prompt_embeds[0]
+            prompt_embeds = prompt_embeds.hidden_states[-2]
+            bs_embed, seq_len, _ = prompt_embeds.shape
+            prompt_embeds = prompt_embeds.view(bs_embed, seq_len, -1)
+            prompt_embeds_list.append(prompt_embeds)
+
+    prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
+    pooled_prompt_embeds = pooled_prompt_embeds.view(bs_embed, -1)
+    return prompt_embeds, pooled_prompt_embeds
+
+
+def prepare_train_dataset(dataset, accelerator):
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [image.convert("RGB") for image in examples[args.image_column]]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [image.convert("RGB") for image in examples[args.conditioning_image_column]]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+
+        return examples
+
+    with accelerator.main_process_first():
+        dataset = dataset.with_transform(preprocess_train)
+
+    return dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    prompt_ids = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
+
+    add_text_embeds = torch.stack([torch.tensor(example["text_embeds"]) for example in examples])
+    add_time_ids = torch.stack([torch.tensor(example["time_ids"]) for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "prompt_ids": prompt_ids,
+        "unet_added_conditions": {"text_embeds": add_text_embeds, "time_ids": add_time_ids},
+    }
+
+
+def main(args):
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    if torch.backends.mps.is_available() and args.mixed_precision == "bf16":
+        # due to pytorch#99272, MPS does not yet support bfloat16.
+        raise ValueError(
+            "Mixed precision training with bfloat16 is not supported on MPS. Please use fp16 (recommended) or fp32 instead."
+        )
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Disable AMP for MPS.
+    if torch.backends.mps.is_available():
+        accelerator.native_amp = False
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizers
+    tokenizer_one = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer",
+        revision=args.revision,
+        use_fast=False,
+    )
+    tokenizer_two = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer_2",
+        revision=args.revision,
+        use_fast=False,
+    )
+
+    # import correct text encoder classes
+    text_encoder_cls_one = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision
+    )
+    text_encoder_cls_two = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision, subfolder="text_encoder_2"
+    )
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder_one = text_encoder_cls_one.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    text_encoder_two = text_encoder_cls_two.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant
+    )
+    vae_path = (
+        args.pretrained_model_name_or_path
+        if args.pretrained_vae_model_name_or_path is None
+        else args.pretrained_vae_model_name_or_path
+    )
+    vae = AutoencoderKL.from_pretrained(
+        vae_path,
+        subfolder="vae" if args.pretrained_vae_model_name_or_path is None else None,
+        revision=args.revision,
+        variant=args.variant,
+    )
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision, variant=args.variant
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        controlnet = ControlNetModel.from_unet(unet)
+
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            if accelerator.is_main_process:
+                i = len(weights) - 1
+
+                while len(weights) > 0:
+                    weights.pop()
+                    model = models[i]
+
+                    sub_dir = "controlnet"
+                    model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                    i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    vae.requires_grad_(False)
+    unet.requires_grad_(False)
+    text_encoder_one.requires_grad_(False)
+    text_encoder_two.requires_grad_(False)
+    controlnet.train()
+
+    if args.enable_npu_flash_attention:
+        if is_torch_npu_available():
+            logger.info("npu flash attention enabled.")
+            unet.enable_npu_flash_attention()
+        else:
+            raise ValueError("npu flash attention requires torch_npu extensions and is supported only on npu devices.")
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warning(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+            controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+        unet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, unet and text_encoder to device and cast to weight_dtype
+    # The VAE is in float32 to avoid NaN losses.
+    if args.pretrained_vae_model_name_or_path is not None:
+        vae.to(accelerator.device, dtype=weight_dtype)
+    else:
+        vae.to(accelerator.device, dtype=torch.float32)
+    unet.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_one.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_two.to(accelerator.device, dtype=weight_dtype)
+
+    # Here, we compute not just the text embeddings but also the additional embeddings
+    # needed for the SD XL UNet to operate.
+    def compute_embeddings(batch, proportion_empty_prompts, text_encoders, tokenizers, is_train=True):
+        original_size = (args.resolution, args.resolution)
+        target_size = (args.resolution, args.resolution)
+        crops_coords_top_left = (args.crops_coords_top_left_h, args.crops_coords_top_left_w)
+        prompt_batch = batch[args.caption_column]
+
+        prompt_embeds, pooled_prompt_embeds = encode_prompt(
+            prompt_batch, text_encoders, tokenizers, proportion_empty_prompts, is_train
+        )
+        add_text_embeds = pooled_prompt_embeds
+
+        # Adapted from pipeline.StableDiffusionXLPipeline._get_add_time_ids
+        add_time_ids = list(original_size + crops_coords_top_left + target_size)
+        add_time_ids = torch.tensor([add_time_ids])
+
+        prompt_embeds = prompt_embeds.to(accelerator.device)
+        add_text_embeds = add_text_embeds.to(accelerator.device)
+        add_time_ids = add_time_ids.repeat(len(prompt_batch), 1)
+        add_time_ids = add_time_ids.to(accelerator.device, dtype=prompt_embeds.dtype)
+        unet_added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
+
+        return {"prompt_embeds": prompt_embeds, **unet_added_cond_kwargs}
+
+    # Let's first compute all the embeddings so that we can free up the text encoders
+    # from memory.
+    text_encoders = [text_encoder_one, text_encoder_two]
+    tokenizers = [tokenizer_one, tokenizer_two]
+    train_dataset = get_train_dataset(args, accelerator)
+    compute_embeddings_fn = functools.partial(
+        compute_embeddings,
+        text_encoders=text_encoders,
+        tokenizers=tokenizers,
+        proportion_empty_prompts=args.proportion_empty_prompts,
+    )
+    with accelerator.main_process_first():
+        from datasets.fingerprint import Hasher
+
+        # fingerprint used by the cache for the other processes to load the result
+        # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
+        new_fingerprint = Hasher.hash(args)
+        train_dataset = train_dataset.map(compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint)
+
+    del text_encoders, tokenizers
+    gc.collect()
+    torch.cuda.empty_cache()
+
+    # Then get the training dataset ready to be passed to the dataloader.
+    train_dataset = prepare_train_dataset(train_dataset, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    # Check the PR https://github.com/huggingface/diffusers/pull/8312 for detailed explanation.
+    num_warmup_steps_for_scheduler = args.lr_warmup_steps * accelerator.num_processes
+    if args.max_train_steps is None:
+        len_train_dataloader_after_sharding = math.ceil(len(train_dataloader) / accelerator.num_processes)
+        num_update_steps_per_epoch = math.ceil(len_train_dataloader_after_sharding / args.gradient_accumulation_steps)
+        num_training_steps_for_scheduler = (
+            args.num_train_epochs * num_update_steps_per_epoch * accelerator.num_processes
+        )
+    else:
+        num_training_steps_for_scheduler = args.max_train_steps * accelerator.num_processes
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=num_warmup_steps_for_scheduler,
+        num_training_steps=num_training_steps_for_scheduler,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        if num_training_steps_for_scheduler != args.max_train_steps * accelerator.num_processes:
+            logger.warning(
+                f"The length of the 'train_dataloader' after 'accelerator.prepare' ({len(train_dataloader)}) does not match "
+                f"the expected length ({len_train_dataloader_after_sharding}) when the learning rate scheduler was created. "
+                f"This inconsistency may result in the learning rate scheduler not functioning properly."
+            )
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                if args.pretrained_vae_model_name_or_path is not None:
+                    pixel_values = batch["pixel_values"].to(dtype=weight_dtype)
+                else:
+                    pixel_values = batch["pixel_values"]
+                latents = vae.encode(pixel_values).latent_dist.sample()
+                latents = latents * vae.config.scaling_factor
+                if args.pretrained_vae_model_name_or_path is None:
+                    latents = latents.to(weight_dtype)
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+
+                # Sample a random timestep for each image
+                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents.float(), noise.float(), timesteps).to(
+                    dtype=weight_dtype
+                )
+
+                # ControlNet conditioning.
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+                down_block_res_samples, mid_block_res_sample = controlnet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=batch["prompt_ids"],
+                    added_cond_kwargs=batch["unet_added_conditions"],
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )
+
+                # Predict the noise residual
+                model_pred = unet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=batch["prompt_ids"],
+                    added_cond_kwargs=batch["unet_added_conditions"],
+                    down_block_additional_residuals=[
+                        sample.to(dtype=weight_dtype) for sample in down_block_res_samples
+                    ],
+                    mid_block_additional_residual=mid_block_res_sample.to(dtype=weight_dtype),
+                    return_dict=False,
+                )[0]
+
+                # Get the target for loss depending on the prediction type
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    target = noise
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                else:
+                    raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+                loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                # DeepSpeed requires saving weights on every device; saving weights only on the main process would cause issues.
+                if accelerator.distributed_type == DistributedType.DEEPSPEED or accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae=vae,
+                            unet=unet,
+                            controlnet=controlnet,
+                            args=args,
+                            accelerator=accelerator,
+                            weight_dtype=weight_dtype,
+                            step=global_step,
+                        )
+
+            logs = {"loss": loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            progress_bar.set_postfix(**logs)
+            accelerator.log(logs, step=global_step)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        # Run a final round of validation.
+        # Setting `vae`, `unet`, and `controlnet` to None to load automatically from `args.output_dir`.
+        image_logs = None
+        if args.validation_prompt is not None:
+            image_logs = log_validation(
+                vae=None,
+                unet=None,
+                controlnet=None,
+                args=args,
+                accelerator=accelerator,
+                weight_dtype=weight_dtype,
+                step=global_step,
+                is_final_validation=True,
+            )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)