【Pycharm】如何在pycharm中调试deepspeed，调试时常见的坑

目录

• 如何在pycharm中调试deepspeed
• 常见的坑
• • 1. valueerror: no slot “”specified on host 'localhost'
  • 2. RuntimeError: The server socket has failed to listen on any local network address. The server socket has failed to bind to [::]:29500

    如何在pycharm中调试deepspeed

    请参考这篇文章：如何在pycharm中调试deepspeed

    （）

    常见的坑

    1. valueerror: no slot “”specified on host ‘localhost’

    deepspeed 单机多卡的时候不要设置export CUDA_VISIBLE_DEVICES，否则容易出现序号的错误。例如设置export CUDA_VISIBLE_DEVICES=4,5 则4,5在localhost的list里成为0,1 如果–include="localhost:4,5"会报错。所以直接设置–include="localhost:4,5"即可1。

    2. RuntimeError: The server socket has failed to listen on any local network address. The server socket has failed to bind to [::]:29500

    注意在py文件前加上这些

    （）

    --include localhost:0 --master_port 5600 --hostfile hostfile_single

    ---

    1. valueerror: no slot “”specified on host ‘localhost’ ↩︎