MedicalGPT

Features

• The first stage: PT (Continue PreTraining) incremental pre-training, pre-training the GPT model twice on massive domain document data to inject domain knowledge
• The second stage: SFT (Supervised Fine-tuning) has supervised fine-tuning, constructs an instruction fine-tuning data set, and performs instruction fine-tuning on the basis of the pre-training model to align instruction intentions
• The third stage: RM (Reward Model) reward model modeling, constructing a human preference ranking data set, training the reward model to align human preferences, mainly the "HHH" principle, specifically "helpful, honest, harmless"
• The fourth stage: RL (Reinforcement Learning) is based on human feedback reinforcement learning (RLHF), using the reward model to train the SFT model, and the generation model uses rewards or penalties to update its strategy in order to generate higher quality, more in line with human preferences
• We provide a simple Gradio-based interactive web interface
• After the training is complete, now we load the trained model to verify the effect of the model generating text

Follow MedicalGPT

MedicalGPT Web Site