File size: 2,165 Bytes
0f36197 0685e69 0f36197 07db590 0f36197 72d12b0 0f36197 2a38d88 93f53be |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 |
---
license: apache-2.0
datasets:
- BAAI/IndustryInstruction_Health-Medicine
- BAAI/IndustryInstruction
base_model:
- MonteXiaofeng/CareBot_Medical_multi-llama3-8b-base
tags:
- 医疗对话模型
- 中英文多语种医疗对话模型
- chatmodel
---
This model is trained from the model: MonteXiaofeng/CareBot_Medical_multi-llama3-8b-base, training data is: BAAI/IndustryInstruction_Health-Medicine， To enhance the model's ability to follow medical instructions and better adapt to specific medical scenarios, we conduct the supervised fine-tuning. This process involves using conversational-style data (comprising both queries and responses) to finetune the pretrained LLM. In the following sections, we will explore the details of data construction and training methods.
## Data Construction
Our SFT dataset comprises a diverse array of question types, including multiple-choice questions from medical exams, single-turn disease diagnoses, and multi-turn health consultations. It integrates data from seven publicly available sources: Chinese Medical Dialogue Data\footnote{https://github.com/Toyhom/Chinese-medical-dialogue-data}, Huatuo26M , MedDialog , ChatMed Consult Dataset , ChatDoctor , CMB\footnote{https://github.com/FreedomIntelligence/CMB}, and MedQA . We preserve portions of authentic doctor-patient conversations and augment the dataset by rewriting the remaining content. For these rewrites, we use real-world medical scenarios as prompts and generate responses via GPT-4. We believe this ensures the diversity of the SFT dataset, which can help the CareBot better adapt to different types of medical problems and patient situations, thereby improving its performance in a variety of scenarios.
## evaluation
evaluation on benchmark is bellow.
gsb result with other medical LLMS
|