RuntimeError: The server socket has failed to listen on any local network address. The server socket

最新推荐文章于 2024-07-25 10:40:32 发布

hc0112

最新推荐文章于 2024-07-25 10:40:32 发布

阅读量1.3w

点赞数 26

本文链接：https://blog.csdn.net/hc0112/article/details/126186341

版权

PyTorch DistributedDataParallel 训练模型地址冲突 master_port

关键词由CSDN通过智能技术生成

Error details: RuntimeError: The server socket has failed to listen on any local network address. The server socket has failed to bind to [::]:29500 (errno: 98 - Address already in use). The server socket has failed to bind to ?UNKNOWN? (errno: 98 - Address already in use).
This error occurs when using torch.nn.parallel.DistributedDataParallel to train a model parallelly. I launched program A with python -m torch.distributed.launch --nproc_per_node=2 trainA.py and worked fine. Then when A is running, I tried to launch program B with python -m torch.distributed.launch --nproc_per_node=2 trainB.py yet ended up with the error above.
It turns out that the issue arises from the network address. As the error reports, the address 29500 is being used. Hence, modifying the address should work. So I used the command python -m torch.distributed.launch --nproc_per_node=2 --master_port='29501' trainB.py.
Problem solved!!!