Instructions to use AesSedai/Step-3.7-Flash-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use AesSedai/Step-3.7-Flash-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="AesSedai/Step-3.7-Flash-GGUF",
	filename="IQ2_S/Step-3.7-Flash-IQ2_S-00001-of-00003.gguf",
)

llm.create_chat_completion(
	messages = "No input example has been defined for this model task."
)

Notebooks
Google Colab
Kaggle
Local Apps Settings

llama.cpp

How to use AesSedai/Step-3.7-Flash-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
llama-cli -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
./llama-cli -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
# Run inference directly in the terminal:
./build/bin/llama-cli -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Use Docker

docker model run hf.co/AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

LM Studio
Jan
Ollama
How to use AesSedai/Step-3.7-Flash-GGUF with Ollama:
```
ollama run hf.co/AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
```

Unsloth Studio

How to use AesSedai/Step-3.7-Flash-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for AesSedai/Step-3.7-Flash-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for AesSedai/Step-3.7-Flash-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for AesSedai/Step-3.7-Flash-GGUF to start chatting

How to use AesSedai/Step-3.7-Flash-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "AesSedai/Step-3.7-Flash-GGUF:Q4_K_M"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use AesSedai/Step-3.7-Flash-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Run Hermes

hermes

Docker Model Runner
How to use AesSedai/Step-3.7-Flash-GGUF with Docker Model Runner:
```
docker model run hf.co/AesSedai/Step-3.7-Flash-GGUF:Q4_K_M
```

Lemonade

How to use AesSedai/Step-3.7-Flash-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull AesSedai/Step-3.7-Flash-GGUF:Q4_K_M

Run and chat with the model

lemonade run user.Step-3.7-Flash-GGUF-Q4_K_M

List all available models

lemonade list

what llama.cpp fork can the IQ4_XS model be run? stepfun's fork is not working.

by rosspanda0 - opened 7 days ago

Discussion

rosspanda0

7 days ago

0.04.895.806 D create_tensor: loading tensor blk.44.exp_probs_b.bias
0.04.895.819 D create_tensor: loading tensor blk.44.ffn_gate_shexp.weight
0.04.895.829 D create_tensor: loading tensor blk.44.ffn_up_shexp.weight
0.04.895.839 D create_tensor: loading tensor blk.44.ffn_down_shexp.weight
0.04.898.746 D done_getting_tensors: tensor 'token_embd.weight' (q8_0) (and 0 others) cannot be used with preferred buffer type CUDA_Host, using CPU instead

AesSedai

Owner 7 days ago

I made the quants from StepFun's fork and loaded it for the PPL/KLD testing, not sure why it wouldn't load?

thirteenbit

6 days ago

Why fork?

Q4_K_M works fine in mainline, just a normal llama.cpp? (although it's sloooow on my machine but that's my hardware - I'm running it on a single RTX3090 with moe CPU offload).

Looking at https://github.com/stepfun-ai/llama.cpp - it's just a python script changes for Step 3.7 gguf conversion?

It's the same code as is in PR here: https://github.com/ggml-org/llama.cpp/pull/23845, yes, only python changes. As long as we have recent llama.cpp and correct GGUF it should just work as well as Step 3.5 + Step 3 VL worked if I understand correctly.

@rosspanda0 , just in case - could you check the sha256 hashes for your downloaded gguf-s?

AesSedai

Owner 6 days ago

Sorry, I said fork but I mean branch. It was late at night when I replied :)

thirteenbit

6 days ago

Fork and IQ4_XS - it was mentioned in the OP's question title, that's why I thought that there's something special with Step 3.7 version from llama.cpp viewpoint, started checking the llama.cpp PR and seen that model architecture is actually supported for a long time - the only changes were in gguf converter.

By the way, @AesSedai , thanks for your amazing work.

Having a model that feels like frontier-level from a few months ago running, even slowly, on a normal desktop feels amazing (well, normal desktop for 2024, in 2026 I'd probably not add 128Gb RAM to a desktop build without thinking, between selecting fan and PSU).

Upload images, audio, and videos by dragging in the text input, pasting, or clicking here.

Tap or paste here to upload images

· Sign up or log in to comment