# shibing624/MedicalGPT

MedicalGPT: Training Your Own Medical GPT Model with ChatGPT Training Pipeline. 训练医疗大模型，实现了包括增量预训练(PT)、有监督微调(SFT)、RLHF、DPO、ORPO、GRPO。

Repository: https://github.com/shibing624/MedicalGPT
Canonical: https://ross.abutalabs.com/products/medicalgpt
Language: Python
License: Apache-2.0
License Family: permissive
Topics: llama, chatgpt, gpt, llm, medical, dpo, medicalgpt
Last push: 2026-06-03T03:39:57+00:00

## Health v2 (maintenance only)
Score: 83/100 (v2, computed 2026-09-02T17:46:02.011165+00:00)
- activity 85, release rhythm 80, longevity 84
- inputs: {"age_days": 1188, "days_push": 91, "days_rel": 135, "gap_med": 6, "n_releases_24m": 4}
- flags: none
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 5748, forks 790 (observed 2026-08-28T04:09:28.988435+00:00)

## What it is
MedicalGPT is a Python training framework for building medical-domain large language models using the full ChatGPT-style training pipeline: incremental pretraining, supervised finetuning, RLHF, DPO, ORPO, GRPO, and on-policy distillation. It supports LoRA and full-parameter training of models like Qwen and LLaMA with DeepSpeed, and includes medical data samples and model releases on Hugging Face.

## Use cases
- train a medical domain LLM from a base model
- finetune an LLM with SFT on medical QA data
- align a chat model with RLHF or DPO
- run GRPO preference optimization training
- distill a large model into a smaller one with on-policy distillation
- finetune a model for tool calling / function call
- build a healthcare chatbot model

## When to choose
- you want a complete, educational pretraining-to-alignment pipeline in one repo
- you need medical-domain LLM finetuning with ready sample datasets
- you want support for many alignment methods (DPO, ORPO, GRPO, OPD) in a single codebase

## When to avoid
- you only need inference or a ready-made medical chatbot without training
- you need a production-grade multi-node training platform with heavy orchestration
- your domain is not medical and you prefer a more general finetuning framework

## Facets
- artifact type: library
- maturity: active
- function: llm-training, machine-learning, deep-learning, rag, chatbot
- domain: large-language-models, healthcare, machine-learning
- platform: python
- tags: llm-finetuning, rlhf, dpo, orpo, grpo, sft, pretraining, medical-ai, huggingface, lora, natural-language-processing, gpu, linux

## Member repositories
- shibing624/MedicalGPT (main) score 83

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:09:28.988435+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-29T17:53:19.084276+00:00, confidence not recorded.
  - readme: https://github.com/shibing624/MedicalGPT (fetched 2026-08-28T04:09:28.988435+00:00, sha 14bdac7e517c)
- Data as of 2026-08-30T08:39:29.467469+00:00.