windows单机单卡训练报错RuntimeError: Distributed package doesn't have NCCL built in

OFA-Sys / Chinese-CLIP

Chinese version of CLIP which achieves Chinese cross-modal retrieval and representation generation.

MIT License

4.01k stars 418 forks source link

windows单机单卡训练报错RuntimeError: Distributed package doesn't have NCCL built in #279

Open bowenzc opened 3 months ago

bowenzc commented 3 months ago

RuntimeError: Distributed package doesn't have NCCL built in

bowenzc commented 3 months ago

EOFError: Ran out of input ERROR:torch.distributed.elastic.multiprocessing.api:failed (exitcode: 1) local_rank: 0 (pid: 27316) of binary: C:\anaconda\envs\clip\python.exe

DtYXs commented 3 months ago

您好，可能是由于Windows系统不支持NCCL分布式后端，可以参考 https://github.com/OFA-Sys/Chinese-CLIP/issues/57#issuecomment-1439787723 尝试将分布式后端设置为gloo看看，更建议在Linux系统训练