deepseek-ai / deepseek-ai/DeepEP

test_low_latency by P2P with 8 bytes width num_token will hang

Open
#548 0 comments 0 reactions 0 assignees View on GitHub
Dominant language
Cuda
Stars
10.1k
Forks
1.4k
Avg merge
4d 1h
Merged PRs (30d)
2

Description

I want test low latency with 8 bytes width num_token, modify deep_ep/config.cpp
struct LowLatencyLayout {

// Symmetric signaling buffers

//size_t dispatch_recv_count_buffer_bytes = num_experts * sizeof(int);
size_t dispatch_recv_count_buffer_bytes = num_experts * sizeof(int64_t);
}

2 H20 and 2 NIC in one node , use cmd start test:
python tests/test_low_latency.py --num-processes 2 --num-topk 2 --num-experts 8

will hang in combine

Contributor guide

No contributing guide indexed for this repository

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.