Skip to content

Latest commit

 

History

History
39 lines (32 loc) · 840 Bytes

README.md

File metadata and controls

39 lines (32 loc) · 840 Bytes

img

ChatXRay is a Multimodel chatbot that input chest radiographs and generate medical report. You can find my paper here.

img

Quick-Start

Of course, you need to modify all the path in the repo to suit your own datapath.

Set up

pip install -r requirements.txt

generate finetuning dataset

./script/down_mimic.py
python generate_QWen_json.py

train

python finetune.py

inference

python infer.py         # use ChatXRay to infer
python Qwen_infer.py    # use Qwen-VL to infer
python ./baselines/Fuyu-8b/infer.py # use Fuyu-8b to infer

generate inference result

python test_baseline.py