Multi-card distribution training problem
Opened this issue · 0 comments
fmk345 commented
When I set up multi-card training, I set the number of GPUs to 2. I did not change other contents in the .sh file, and the following error was reported: "ValueError: Expected a string path to an existing deepspeed config, or a dictionary, or a base64 encoded string. Received: stage3_no_offloading_accelerate.conf"