Search results

Connections time out

Status: Open.

#21 In tczhangzhi/pytorch-distributed;

· ShuGuoJ opened on Nov 4, 2023

Can't pickle <function main_worker

Status: Open.

#20 In tczhangzhi/pytorch-distributed;

· Asunatan opened on Dec 4, 2022

RuntimeError: Timed out initializing process group in store based barrier on rank: 2, for key: store_based_barrier_key:1 (world_size=2, worker_count=4, timeout=0:30:00)

Status: Open.

#18 In tczhangzhi/pytorch-distributed;

· poetryben88 opened on Jan 6, 2022

how can i modify it ?

Status: Open.

#17 In tczhangzhi/pytorch-distributed;

· FANG-MING opened on Sep 6, 2021

how can i modify it

Status: Open.

#16 In tczhangzhi/pytorch-distributed;

· FANG-MING opened on Sep 6, 2021

Windows 提示Distributed package doesn't have NCCL "Distributed package doesn't have NCCL built in

Status: Open.

#15 In tczhangzhi/pytorch-distributed;

· Amanda-Qu opened on Aug 4, 2021

多卡显存占用不均衡

Status: Open.

#13 In tczhangzhi/pytorch-distributed;

· MachineVision123 opened on Mar 1, 2021

您好，我想请问一下Pytorch 在多进程分布式训练的时候，一开始加载数据集dataloader的时候，由于是多进程的，就会导致多个进程一起加载dataloader，这些dataloder全部加载到内存中使得内存爆炸，请问有没有什么办法解决呢

Status: Open.

#12 In tczhangzhi/pytorch-distributed;

· 1456416403 opened on Feb 23, 2021

大佬，请问如何指定gpu训练

Status: Open.

#10 In tczhangzhi/pytorch-distributed;

· ccijunk opened on Feb 4, 2021

请问使用DistributedSampler，各个GPU的数据是如何分配的？是连续(互不相同)的还是相同的？

Status: Open.

#9 In tczhangzhi/pytorch-distributed;

· fhong-jpg opened on Jan 6, 2021

关于损失backward问题

Status: Open.

#8 In tczhangzhi/pytorch-distributed;

· menghuanlater opened on Dec 21, 2020

[Discussion] mp: duplicate of torch.cuda.set_device(local_rank) and images = images.cuda(local_rank, non_blocking=True)

enhancementNew feature or requestNew feature or request

Status: Open.

#5 In tczhangzhi/pytorch-distributed;

· laoreja opened on Jun 26, 2020