Search results

Cyclical Latency Variation in <code>combine</code> Operation During Unbalanced MoE Inference

Status: Open.

#485 In deepseek-ai/DeepEP;

· FirwoodLin opened on Nov 6, 2025

Questions about the Necessity of NVIDIA Driver Config Parameters

Status: Open.

#484 In deepseek-ai/DeepEP;

· michaelchen1996 opened on Nov 6, 2025

How to decide the quant group scale

Status: Open.

#482 In deepseek-ai/DeepEP;

· DDDVE opened on Nov 5, 2025

Does DeepEP support sm86?(4A6000 without NVL)

Status: Open.

#481 In deepseek-ai/DeepEP;

· ziyuhuang123 opened on Nov 5, 2025

[bug]Missing some critical error log when timeout

Status: Open.

#480 In deepseek-ai/DeepEP;

· huakainomore opened on Nov 5, 2025

cannot run internode test on 2 Node setup , with each having Two GPU - num_ranks &gt; NUM_MAX_NVL_PEERS or low_latency_mode'

Status: Open.

#477 In deepseek-ai/DeepEP;

· alokprasad opened on Nov 3, 2025

Question: asymmetric-domain bandwidth forwarding

Status: Open.

#475 In deepseek-ai/DeepEP;

· alokprasad opened on Oct 31, 2025

<code>dispatch_wait_recv_cost_stats</code> is cumulating time from each warp, when instead computing a max across warps would be more helpful

Status: Open.

#473 In deepseek-ai/DeepEP;

· goelayu opened on Oct 30, 2025

Ask about the difference of bandwidth between dispatch and. combine

Status: Open.

#472 In deepseek-ai/DeepEP;

· kzlxd opened on Oct 30, 2025

Questions About AR Support for Low-latency

Status: Open.

#467 In deepseek-ai/DeepEP;

· michaelchen1996 opened on Oct 27, 2025

Can deep_ep run on environments with more than 160 ranks?

Status: Open.

#464 In deepseek-ai/DeepEP;

· ZhenguoYao1 opened on Oct 21, 2025

Does deepEP support PCIe H800s？

Status: Open.

#462 In deepseek-ai/DeepEP;

· Weigaa opened on Oct 20, 2025