PaddleNLP
PaddleNLP copied to clipboard
support multicard export for qwen
PR types
New features
PR changes
Others
Description
support multicard export for qwen
ιͺθ―ζΉζ³οΌ θΏθ‘δΈι’ηε½δ»€θΏθ‘δΈ€ε‘ε―ΌεΊ:
python -m paddle.distributed.launch --gpus 1,2 export_model.py --model_name_or_path qwen-vl/qwen-vl-7b-static --output_path ./QWen_check_points --dtype float16 --inference_model