Instructions to use epfl-llm/meditron-70b with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use epfl-llm/meditron-70b with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="epfl-llm/meditron-70b")# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("epfl-llm/meditron-70b") model = AutoModelForCausalLM.from_pretrained("epfl-llm/meditron-70b", device_map="auto") - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use epfl-llm/meditron-70b with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "epfl-llm/meditron-70b" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "epfl-llm/meditron-70b", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'Use Docker
docker model run hf.co/epfl-llm/meditron-70b
- SGLang
How to use epfl-llm/meditron-70b 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 "epfl-llm/meditron-70b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "epfl-llm/meditron-70b", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }'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 "epfl-llm/meditron-70b" \ --host 0.0.0.0 \ --port 30000 # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:30000/v1/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "epfl-llm/meditron-70b", "prompt": "Once upon a time,", "max_tokens": 512, "temperature": 0.5 }' - Docker Model Runner
How to use epfl-llm/meditron-70b with Docker Model Runner:
docker model run hf.co/epfl-llm/meditron-70b
The Development Concept of This Model is So Cool
I am a Chinese native speaker and a passionate AI enthusiast. I really appreciate the development concept of democratizing healthcare. I believe this project will benefit everyone, especially in the context of the unequal distribution of medical resources. With a high-configuration computer and Meditron:7b, valuable medical reference information can be provided in any corner of the world, helping thousands of people. I mainly tested the 7b version in English (because I believe the lightweight nature of this AI will benefit the masses), and I am not a professional programmer; my AI knowledge comes from related books. Here are some issues I encountered during testing that need improvement:
Meditron:7b's response length is limited, which often leads to a lack of precision in its answers, a very common issue.
Meditron:7b seems unable to recognize some drug code names that were in testing before August 2023.
Meditron:7b is not clear in explaining the pharmacology of newer drugs like Neflamapimod.
Its ability to provide medical advice in cases of multiple concurrent diseases is somewhat weak.
However, what surprised me was its good performance in identifying rare diseases and complex medical indicators, showcasing excellent medical reasoning abilities. I really hope to contribute to this project in any way I can, such as providing Chinese language testing for future versions!