Upload folder using huggingface_hub

checkpoint_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6bb07ec05bd4f4c49eb41d351e0c3b736eeb286e57c619acbbf0d9d4ba2c9aa
+size 2897697355

checkpoint_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d22cc65257192dcf16a1367c3d3e903387a5218d8dc18b14a133543c283105b
+size 2897697355

checkpoint_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df666a589e7e218392670d8db75007ff53015645e836bd8272dc9d84fbd2958a
+size 2897697355

checkpoint_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1b2d7f7ac32b75c0e79623f4430ea3f167e27bc5b904baaba703a68a9b940ab
+size 2897697355

checkpoint_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:506125389afa9f0ec9161856146e0f7814c21f7064b3270be38f99a32be83e4c
+size 2897697355

checkpoint_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9d6d562e5ebf4e74e76c6bf816d04d9b397a69ef4741e6c4655f616053a3110
+size 2897697355

log.txt

@@ -0,0 +1,118 @@
+{
+    "run": {
+        "task": "video_text_pretrain",
+        "lr_sched": "linear_warmup_cosine_lr",
+        "init_lr": 0.0001,
+        "min_lr": 1e-05,
+        "warmup_lr": 1e-06,
+        "weight_decay": 0.05,
+        "max_epoch": 6,
+        "iters_per_epoch": 3102,
+        "batch_size_train": 1,
+        "batch_size_eval": 4,
+        "num_workers": 4,
+        "warmup_steps": 3102,
+        "accum_grad_iters": 8,
+        "seed": 42,
+        "output_dir": "ckpt/timechat/train_stage2_llama2_7b_instruct12.4k_charades_bz4_f96_epoch6_ws32_stride32_mfp96_mtl2048_lr1e-4",
+        "amp": true,
+        "resume_ckpt_path": null,
+        "evaluate": false,
+        "train_splits": [
+            "train"
+        ],
+        "device": "cuda",
+        "world_size": 4,
+        "dist_url": "env://",
+        "distributed": true,
+        "rank": 0,
+        "gpu": 0,
+        "dist_backend": "nccl"
+    },
+    "model": {
+        "arch": "timechat",
+        "image_size": 224,
+        "drop_path_rate": 0,
+        "use_grad_checkpoint": true,
+        "vit_precision": "fp16",
+        "freeze_vit": true,
+        "freeze_qformer": false,
+        "num_query_token": 32,
+        "llama_model": "ckpt/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+        "prompt": "",
+        "model_type": "pretrain_llama_v2",
+        "vit_model": "ckpt/eva-vit-g/eva_vit_g.pth",
+        "q_former_model": "ckpt/instruct-blip/instruct_blip_vicuna7b_trimmed.pth",
+        "ckpt": "ckpt/timechat/timechat_7b.pth",
+        "frozen_llama_proj": false,
+        "frozen_video_Qformer": false,
+        "fusion_head_layers": 2,
+        "max_frame_pos": 96,
+        "fusion_header_type": "seqTransf",
+        "max_txt_len": 2048,
+        "end_sym": "</s>",
+        "prompt_path": "",
+        "prompt_template": "[INST] <<SYS>>\\n \\n<</SYS>>\\n\\n{} [/INST] ",
+        "lora": true,
+        "lora_inference_mode": false,
+        "qformer_text_input": true,
+        "window_size": 32,
+        "stride": 32
+    },
+    "preprocess": {
+        "vis_processor": {
+            "train": {
+                "name": "alpro_video_train",
+                "image_size": 224,
+                "n_frms": 8
+            },
+            "eval": {
+                "name": "alpro_video_eval",
+                "image_size": 224,
+                "n_frms": 8
+            }
+        },
+        "text_processor": {
+            "train": {
+                "name": "blip_caption"
+            },
+            "eval": {
+                "name": "blip_caption"
+            }
+        }
+    },
+    "datasets": {
+        "charades_instruct": {
+            "data_type": "video",
+            "build_info": {
+                "anno_dir": "data/TimeIT/data/temporal_video_grounding/charades/instruct_tvg_12.4k_charades.json",
+                "videos_dir": "data/"
+            },
+            "vis_processor": {
+                "train": {
+                    "name": "alpro_video_train",
+                    "n_frms": 96,
+                    "image_size": 224
+                }
+            },
+            "text_processor": {
+                "train": {
+                    "name": "blip_caption"
+                }
+            },
+            "num_video_query_token": 32,
+            "tokenizer_name": "ckpt/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+            "model_type": "llama_v2",
+            "num_frm": 96,
+            "sample_type": "rand",
+            "max_txt_len": 2048,
+            "stride": 32
+        }
+    }
+}
+{"train_lr": "0.000", "train_loss": "0.482"}
+{"train_lr": "0.000", "train_loss": "0.443"}
+{"train_lr": "0.000", "train_loss": "0.426"}
+{"train_lr": "0.000", "train_loss": "0.409"}
+{"train_lr": "0.000", "train_loss": "0.387"}
+{"train_lr": "0.000", "train_loss": "0.361"}

tvg/fmt_charades_test_f96_result.txt

@@ -0,0 +1,4 @@
+# pred video timestamps 3720; # gt video timestamps 3720
+IOU 0.3: 70.3763440860215
+IOU 0.5: 53.38709677419355
+IOU 0.7: 28.252688172043012

tvg/log.txt

@@ -0,0 +1,112 @@
+{
+    "run": {
+        "task": "video_text_pretrain"
+    },
+    "model": {
+        "arch": "timechat",
+        "image_size": 224,
+        "drop_path_rate": 0,
+        "use_grad_checkpoint": true,
+        "vit_precision": "fp16",
+        "freeze_vit": true,
+        "freeze_qformer": true,
+        "num_query_token": 32,
+        "llama_model": "ckpt/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+        "prompt": "",
+        "model_type": "pretrain_llama_v2",
+        "max_txt_len": 2048,
+        "end_sym": "</s>",
+        "low_resource": false,
+        "frozen_llama_proj": true,
+        "frozen_video_Qformer": true,
+        "vit_model": "ckpt/eva-vit-g/eva_vit_g.pth",
+        "q_former_model": "ckpt/instruct-blip/instruct_blip_vicuna7b_trimmed.pth",
+        "ckpt": "timechat/ckpt/timechat/train_stage2_llama2_7b_instruct12.4k_charades_bz4_f96_epoch6_ws32_stride32_mfp96_mtl2048_lr1e-4/20240407024/checkpoint_5.pth",
+        "fusion_head_layers": 2,
+        "max_frame_pos": 96,
+        "fusion_header_type": "seqTransf",
+        "lora": true,
+        "lora_inference_mode": true,
+        "qformer_text_input": true,
+        "window_size": 32,
+        "stride": 32,
+        "device_8bit": "0"
+    },
+    "preprocess": {
+        "vis_processor": {
+            "train": {
+                "name": "alpro_video_train",
+                "image_size": 224,
+                "n_frms": 8
+            },
+            "eval": {
+                "name": "alpro_video_eval",
+                "image_size": 224,
+                "n_frms": 8
+            }
+        },
+        "text_processor": {
+            "train": {
+                "name": "blip_caption"
+            },
+            "eval": {
+                "name": "blip_caption"
+            }
+        }
+    },
+    "datasets": {
+        "webvid": {
+            "data_type": "video",
+            "build_info": {
+                "anno_dir": "path/webvid/webvid_tain_data/annotations/",
+                "videos_dir": "path//webvid/webvid_tain_data/videos/"
+            },
+            "vis_processor": {
+                "train": {
+                    "name": "alpro_video_eval",
+                    "n_frms": 96,
+                    "image_size": 224
+                }
+            },
+            "text_processor": {
+                "train": {
+                    "name": "blip_caption"
+                }
+            },
+            "num_video_query_token": 32,
+            "tokenizer_name": "ckpt/Video-LLaMA-2-7B-Finetuned/llama-2-7b-chat-hf/",
+            "model_type": "llama_v2",
+            "num_frm": 96,
+            "sample_type": "uniform",
+            "max_txt_len": 2048,
+            "stride": 32
+        }
+    }
+}
+
+cfg_path                 : eval_configs/timechat.yaml
+anno_path                : data/TimeIT/data/temporal_video_grounding/charades/charades_annotation
+video_path               : data/Charades/videos/
+model_type               : None
+task                     : tvg
+dataset                  : charades
+output_dir               : timechat/ckpt/timechat/train_stage2_llama2_7b_instruct12.4k_charades_bz4_f96_epoch6_ws32_stride32_mfp96_mtl2048_lr1e-4/20240407024/tvg
+split                    : test
+num_frames               : 96
+top_p                    : 0.8
+temperature              : 1
+batch_size               : 8
+gpu_id                   : 0
+timestamp                : False
+timestamp_file           : 
+debug                    : False
+prompt_file              : prompts/tvg_description_zeroshot.txt
+timechat_model_path      : timechat/ckpt/timechat/train_stage2_llama2_7b_instruct12.4k_charades_bz4_f96_epoch6_ws32_stride32_mfp96_mtl2048_lr1e-4/20240407024/checkpoint_5.pth
+sample_num               : -1
+example_output           : False
+no_lora                  : False
+post_check               : False
+post_check_prompt_file   : prompts/dvc_post_check.txt
+asr                      : False
+asr_path                 : data/YouCook2-BB/YouCook2_asr_denseCap/whisper_outputs_with_time/small.en.cleaned/
+options                  : []