mirror of
https://github.com/deepseek-ai/DeepEP.git
synced 2026-11-05 00:47:32 +00:00
[GH-ISSUE #495] A800 adapt @898269fa06 #320
Labels
No labels
pull-request
No milestone
No project
No assignees
1 participant
Notifications
Due date
No due date set.
Dependencies
No dependencies set.
Reference
deepseek-ai/DeepEP#320
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @kzlxd on GitHub (Nov 14, 2025).
Original GitHub issue: https://github.com/deepseek-ai/DeepEP/issues/495
I have adapted the A800 on
898269fa06, and the TMA-related components have been successfully compiled. However, the following error occurs during runtime. Could you advise how this should be addressed?[2025-11-14 09:45:28] [192.168.80.96] WARN: found inactive port or port with non IB/RoCE link layer protocol, skipping..
/opt/DeepEP/csrc/kernels/internode.cu:372, condition: ibgda_get_state()->num_rc_per_pe == num_channels or ibgda_get_state()->num_rc_per_pe >= num_sms @sphish
@sphish commented on GitHub (Nov 17, 2025):
Could you print
ibgda_get_state()->num_rc_per_pe?@kzlxd commented on GitHub (Nov 17, 2025):
thanks，i fixed it because the number of ibgda_get_state()->num_rc_per_pe is 16. but new error occur at File "/opt/DeepEP/tests/test_new.py", line 132, in test_main
[2025-11-17 10:01:26] [192.168.80.96] assert gbl_num_tokens_per_rank[rank].item() == recv_x.size(0), f'{gbl_num_tokens_per_rank[rank].item()} != {recv_x.size(0)}', it seems to cant access the gbl_num_tokens_per_rank[rank].item().i print it ,it will occur error. what about this error? could you give me some advice? @sphish