Neura Market
ChatDoctor logo

ChatDoctor

Free

3.4k

FreeFree tier
Inputs: textOutputs: text
Type
Open Source
Visit Website GitHub (40 stars)

About ChatDoctor

ChatDoctor is an open-source medical chat model that fine-tunes Meta AI's LLaMA large language model using domain-specific medical knowledge. Developed by researchers from UT Southwestern Medical Center, the University of Illinois at Urbana-Champaign, Ohio State University, and Hangzhou Dianzi University, the model is trained on datasets comprising real conversations between patients and doctors, including 100k interactions from HealthCareMagic.com and 10k from icliniq.com. The project provides a complete pipeline for training, including scripts for fine-tuning with LoRA, and offers an interactive demo hosted on Hugging Face Spaces. It is important to note that the model has not achieved 100% accurate output and is not intended for real clinical use.

Key Features

Fine-tuned on LLaMA using medical domain knowledge
Trained on real patient-doctor conversation datasets (HealthCareMagic-100k, icliniq-10k)
Open-source with training and inference scripts available on GitHub
Includes an interactive demo on Hugging Face Spaces
Supports LoRA (Low-Rank Adaptation) for efficient fine-tuning
Provides a disease database for autonomous consultation capabilities

Pros & Cons

Pros
  • Open-source and freely available for research and non-commercial use
  • Trained on real medical conversations, providing domain-relevant responses
  • Includes multiple datasets and pretrained checkpoints for reproducibility
  • Supports efficient fine-tuning via LoRA, reducing computational requirements
  • Provides an online demo for easy testing without local setup
Cons
  • Model accuracy is not 100% and should not be used in real clinical scenarios
  • Requires access to a GPU and PyTorch environment for local training/inference
  • Free tier of the Hugging Face demo may have usage limits; access requires registration
  • Dependence on LLaMA weights, which have their own licensing restrictions
  • Output quality may vary and should be verified by medical professionals

Best For

Medical question answering and symptom explanationResearch in medical AI and natural language processingEducational demonstrations of LLM fine-tuning in healthcarePrototyping conversational agents for medical triage (non-clinical)Exploring domain-specific adaptation of large language models

FAQ

Is ChatDoctor free to use?
The model and code are open-source and appear to be freely available for research and non-commercial use. The online demo on Hugging Face Spaces may have usage limits; exact terms should be verified on the demo page.
Can ChatDoctor be used for real medical diagnosis?
No. The project explicitly states that the model has not achieved 100% accurate output and should not be applied to real clinical scenarios. It is intended for research and educational purposes only.
What datasets were used to train ChatDoctor?
The model was trained on 100k real conversations between patients and doctors from HealthCareMagic.com and 10k conversations from icliniq.com, along with Stanford Alpaca data for basic conversational capabilities.
How can I run ChatDoctor locally?
You can clone the GitHub repository, set up a conda environment with PyTorch, install dependencies from requirements.txt, and use the provided training and inference scripts. A GPU is recommended.
Does ChatDoctor support languages other than English?
Based on available information, the model appears to be trained primarily on English-language medical conversations. Support for other languages should be verified.