Instructions to use mistralai/Mistral-7B-Instruct-v0.2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use mistralai/Mistral-7B-Instruct-v0.2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="mistralai/Mistral-7B-Instruct-v0.2") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2") model = AutoModelForCausalLM.from_pretrained("mistralai/Mistral-7B-Instruct-v0.2") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Inference
- Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use mistralai/Mistral-7B-Instruct-v0.2 with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Install mistral-common: pip install --upgrade mistral-common # Start the vLLM server: vllm serve "mistralai/Mistral-7B-Instruct-v0.2" --tokenizer_mode mistral --config_format mistral --load_format mistral --tool-call-parser mistral --enable-auto-tool-choice # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "mistralai/Mistral-7B-Instruct-v0.2", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/mistralai/Mistral-7B-Instruct-v0.2
- SGLang
How to use mistralai/Mistral-7B-Instruct-v0.2 with SGLang:
Install from pip and serve model
# Install SGLang from pip: pip install sglang # Start the SGLang server: python3 -m sglang.launch_server \ --model-path "mistralai/Mistral-7B-Instruct-v0.2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "mistralai/Mistral-7B-Instruct-v0.2", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker images
docker run --gpus all \ --shm-size 32g \ -p 30000:30000 \ -v ~/.cache/huggingface:/root/.cache/huggingface \ --env "HF_TOKEN=<secret>" \ --ipc=host \ lmsysorg/sglang:latest \ python3 -m sglang.launch_server \ --model-path "mistralai/Mistral-7B-Instruct-v0.2" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "mistralai/Mistral-7B-Instruct-v0.2", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use mistralai/Mistral-7B-Instruct-v0.2 with Docker Model Runner:
docker model run hf.co/mistralai/Mistral-7B-Instruct-v0.2
Sliding window is null in config file
I notice in config.json that the sliding window is set to null. Does this mean that the defacto sliding window is the full embedding input length of 32k?
FWIW, running Mistral v0.1 there always seemed to be issues running contexts above the sliding window length (e.g. see this issue), so maybe that's related?
The hugging face docs for Mistral say that if set to None, it will default to 4096: https://huggingface.co/docs/transformers/main/en/model_doc/mistral
I am not sure what is the intention here
Digging deeper, setting it to none will indeed disable it: https://github.com/huggingface/transformers/blob/8eb74c1c8961e3dc8549bb1a76463c7658a63d43/src/transformers/models/mistral/modeling_mistral.py#L380
@KurtGD1915 It says under sliding window that it defaults to 4096, but looking at the code, it is just default param value (so if you set it to None explicitly, it will stay none). I have fixed the link above that actually does the inference, where you can see that sliding window is disabled if it's None or larger than content length.