多卡部署提示不在同一个设备:Expected all tensors to be on the same device, but found at least two devices, cuda:2 and cuda:3!

#4
by Winkuis - opened

File "/root/.cache/huggingface/modules/transformers_modules/cogvlm2-llama3-chinese-chat-19B/visual.py", line 83, in forward
output = mlp_input + mlp_output
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:2 and cuda:3!

这问题咋解决,API部署的!

Knowledge Engineering Group (KEG) & Data Mining at Tsinghua University org

单卡部署,否则应该修改代码成 cli_demo_multi_gpus.py这种载入方式

zRzRzRzRzRzRzR changed discussion status to closed

Sign up or log in to comment