[GH-ISSUE #46] zero3单机多卡前向传播会有问题 #41

Open
opened 2026-10-26 21:02:00 +00:00 by bsakaguchi · 0 comments
Owner

Originally created by @SWORD-ZEUS on GitHub (Apr 5, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-MoE/issues/46

我两张卡zero3 deepspeed配置下,前向传播(参数冻结)的时候,decoder_layer经过了两层,就出现了gpu通信问题

Image

Originally created by @SWORD-ZEUS on GitHub (Apr 5, 2025). Original GitHub issue: https://github.com/deepseek-ai/DeepSeek-MoE/issues/46 我两张卡zero3 deepspeed配置下,前向传播(参数冻结)的时候,decoder_layer经过了两层,就出现了gpu通信问题 ![Image](https://github.com/user-attachments/assets/817cac78-a7ea-4ffa-9e2e-65fd241bb1c4)
Sign in to join this conversation.
No labels
pull-request
WIP
No milestone
No project
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepSeek-MoE#41
No description provided.