Skip to content

Issues: deepseek-ai/DeepEP

Easier Potential Overall Design
#39 opened Mar 4, 2025 by LyricZhao
Open
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

28 Open 40 Closed
28 Open 40 Closed
Author
Filter by author
Loading
Label
Filter by label
Loading
Use alt + click/return to exclude labels
or ⇧ + click/return for logical OR
Projects
Filter by project
Loading
Milestones
Filter by milestone
Loading
Assignee
Filter by who’s assigned
Sort
Sort by
Newest Oldest Most commented Least commented Recently updated Least recently updated Best match
Most reactions
👍 👎 😄 🎉 😕 ❤️ 🚀 👀

Issues list

test_internode.py hangs at internode.cu:dispatch(...)
#74 opened Mar 14, 2025 by yanminjia
3
Question about interface of low_latency_dispatch
#72 opened Mar 13, 2025 by JimmyWang1850
1
Confues about the Low-latency kernels data issue
#71 opened Mar 12, 2025 by ysywh
1
About the setting of kNumWarpsPerGroup and kNumWarpGroups
#63 opened Mar 10, 2025 by GitAlice123
1
Questions about Compute-Communication Overlap
#61 opened Mar 9, 2025 by Leaf996
1
RDMA bandwidth is not normal when run test_internode.py.
#60 opened Mar 9, 2025 by MinhuiWan
6
AllToAll performance not improving
#59 opened Mar 8, 2025 by cryoco
3
output tokens in intranode::dispatch for each expert isn't packed?
#58 opened Mar 7, 2025 by zhang662817
1
test_low_latency failed
#55 opened Mar 7, 2025 by hyesung84
1
Why is it possible to disable congestion control?
#52 opened Mar 7, 2025 by JJBM123
3
Confuse about the estimated bandwidth on tests/test_internode.py
#51 opened Mar 6, 2025 by Fangjin98
3
Why NUM_MAX_NVL_PEERS must be 8?
#50 opened Mar 6, 2025 by Huixxi
1
Assertion failed: internode_ll.cu:259, condition: num_recv_tokens != 0
#49 opened Mar 5, 2025 by abcdabcd987
4
How to apply Moe MLP to recv_x returned by dispatch when num_expert_per_rank > 1
#48 opened Mar 5, 2025 by CyCle1024
2
About num_qps_per_rank
#46 opened Mar 5, 2025 by GitAlice123
5
Easier Potential Overall Design
#39 opened Mar 4, 2025 by LyricZhao
Test test_low_latency.py failed on H100 with ROCE
#38 opened Mar 3, 2025 by ImbaPlayer
32
Intra-node compute-communication overlap without SM cores
#37 opened Mar 3, 2025 by rajagond
7
Is setting IBGDA necessary for test_internode.py?
#36 opened Mar 2, 2025 by zhangml
14
Why is a mechanism like PXN in NCCL not used in low-latency kernel? It should be able to achieve even lower latency.
#31 opened Feb 28, 2025 by weishenmezhememan
3
When testing test_internode.py, what is the effect of setting NVSHMEM_DISABLE_P2P=1? Will NVLink be disabled?
#30 opened Feb 28, 2025 by MoringKing
4
"gdrdrv-dkms:amd64 is not configured" in container when installing NVSHMEM
#24 opened Feb 26, 2025 by JoeyYoung
5
Test test_low_latency.py failed on H20 ROCE
#17 opened Feb 25, 2025 by HanHan009527
14
Test test_low_latency.py failed
#15 opened Feb 25, 2025 by Nekofish-L
5
Test fail use 2node on tests/test_internode.py
#13 opened Feb 25, 2025 by xesdiny
8
ProTip! Find all open issues with in progress development work with linked:pr.