DeepSeek/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP synced 2025-05-07 05:24:27 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Merge pull request #89 from fzyzcjy/patch-1

Browse Source 
Super tiny fix typo
This commit is contained in:
Chenggang Zhao 2025-03-25 09:28:44 +08:00 committed by GitHub
commit 7d52ad7248
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
deep_ep/buffer.py
View File

@ -479,7 +479,7 @@ class Buffer:
Moreover, not all tokens are valid, only some of the `num_max_dispatch_tokens_per_rank * num_ranks` are, Moreover, not all tokens are valid, only some of the `num_max_dispatch_tokens_per_rank * num_ranks` are,
as we do not synchronize CPU received count with GPU (also not incompatible with CUDA graph if synced). as we do not synchronize CPU received count with GPU (also not incompatible with CUDA graph if synced).
recv_count: a tensor shaped `[num_local_experts]` with type `torch.int`, indicating how many tokens each recv_count: a tensor shaped `[num_local_experts]` with type `torch.int`, indicating how many tokens each
expert receive. As mentioned before, all not tokens are valid in `recv_x`. expert receive. As mentioned before, not all tokens are valid in `recv_x`.
handle: the communication handle to be used in the `low_latency_combine` function. handle: the communication handle to be used in the `low_latency_combine` function.
event: the event after executing the kernel (valid only if `async_finish` is set). event: the event after executing the kernel (valid only if `async_finish` is set).
hook: the receiving hook function (valid only if `return_recv_hook` is set). hook: the receiving hook function (valid only if `return_recv_hook` is set).