OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Medicalgpt

MedicalGPT is a Python framework in a public GitHub repository for training medical GPT-style language models through continued pretraining, supervised fine-tuning and preference-optimization methods.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Medicalgpt — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you intend to run model training or fine-tuning yourself and need a code-based pipeline for medical-domain adaptation.

Who it is for

  • Developers adapting GPT-style models to medical-domain data.
  • Research teams exploring fine-tuning and alignment workflows for domain language models.

Documented uses

  • Continue pretraining a base model on medical-domain documents.
  • Supervise fine-tune a model for a medical-domain task.
  • Apply preference-optimization approaches such as RLHF, DPO, ORPO or GRPO.
Before you choose

Requirements & limits

The repository is public and identifies its license as Apache-2.0.

  • It is a training framework, not a documented hosted inference API.
  • The stated pipeline includes optional continued pretraining, supervised fine-tuning and several preference-optimization methods.
  • Model-family and method support should be checked against the current repository instructions for the intended workflow.
  • The user must arrange suitable training data and computing resources.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install MedicalGPT dependencies

Clone the repository and install or update its Python dependencies.

Before you start

  • Git and pip are available.
  • The README badge indicates Python 3.8 or later.

Steps

  1. Clone the MedicalGPT repository: the command below.

    Documented command
    git clone https://github.com/shibing624/MedicalGPT

  2. Change into the repository directory: cd MedicalGPT.

  3. Install or update the listed dependencies: the command below.

    Documented command
    pip install -r requirements.txt --upgrade

Expected outcome

The repository is cloned locally and dependencies from requirements.txt are installed or upgraded.

Guide limitations

  • This covers dependency setup only, not training or inference.
  • The README notes that requirements.txt may change over time.
  • The README lists estimated GPU memory needs that vary by model size, training method, and precision.
Documentation 1 ↗
Source-based practical guide

Run interactive inference with a base model and optional LoRA adapter

Start interactive text generation using the repository's inference script.

Before you start

  • A Hugging Face-format base-model directory containing model weights and configuration.
  • A compatible tokenizer directory; by default, the tokenizer path is the same as the base-model path.
  • For adapter-based inference, a LoRA directory or a Hugging Face Model Hub model name; omit it if LoRA weights are already merged into the base model.
  • The repository dependencies are installed.

Steps

  1. From the repository directory, set CUDA_VISIBLE_DEVICES=0 and run the command below with --base_model set to the base-model directory.

    Documented command
    python demo/inference.py

  2. Optionally provide --lora_model with the LoRA directory or Hub model name.

  3. Add --interactive to start multi-turn streaming inference.

Expected outcome

The script starts interactive, multi-turn text generation with the specified base model and optional LoRA adapter.

Guide limitations

  • The example selects GPU 0; the README separately documents --only_cpu and --gpus options.
  • The README describes model and tokenizer paths in Hugging Face format.
  • This recipe does not establish a hosted API or guarantee output quality.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
7 October 2026
Original website title
shibing624/MedicalGPT · Official homepage title · retained in its source language
How we review sourcesSuggest a correction