[PR #743] [CLOSED] Align fp8 quantization #748

New issue

Closed

opened 2026-10-27 21:05:50 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-27 21:05:50 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/743
Author: @ArronHZG
Created: 8/31/2026
Status: ❌ Closed

Base: main ← Head: align_fp8_quantization

📝 Commits (3)

2bd56ae fix: align GLM-5 low-latency dispatch
6845ffd Add align_fp8_quantization
c5c2b0c feat: align low-latency FP8 across SGLang versions

📊 Changes

7 files changed (+142 additions, -22 deletions)

View changed files

📝 csrc/deep_ep.cpp (+4 -1)
📝 csrc/deep_ep.hpp (+2 -1)
📝 csrc/kernels/api.cuh (+1 -0)
📝 csrc/kernels/internode_ll.cu (+80 -12)
📝 deep_ep/buffer.py (+42 -5)
📝 tests/test_internode.py (+2 -2)
📝 tests/test_low_latency.py (+11 -1)

📄 Description

No description provided

_{🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.}

## 📋 Pull Request Information **Original PR:** https://github.com/deepseek-ai/DeepEP/pull/743 **Author:** [@ArronHZG](https://github.com/ArronHZG) **Created:** 8/31/2026 **Status:** ❌ Closed **Base:** `main` ← **Head:** `align_fp8_quantization` --- ### 📝 Commits (3) - [`2bd56ae`](https://github.com/deepseek-ai/DeepEP/commit/2bd56ae16b9c309505fbcc062798834d88ce074d) fix: align GLM-5 low-latency dispatch - [`6845ffd`](https://github.com/deepseek-ai/DeepEP/commit/6845ffd9d59126ec0030c13e0e155935a61e5b5a) Add align_fp8_quantization - [`c5c2b0c`](https://github.com/deepseek-ai/DeepEP/commit/c5c2b0cfa767b1afeb4d237c4ed7e5ff062a550d) feat: align low-latency FP8 across SGLang versions ### 📊 Changes **7 files changed** (+142 additions, -22 deletions) <details> <summary>View changed files</summary> 📝 `csrc/deep_ep.cpp` (+4 -1) 📝 `csrc/deep_ep.hpp` (+2 -1) 📝 `csrc/kernels/api.cuh` (+1 -0) 📝 `csrc/kernels/internode_ll.cu` (+80 -12) 📝 `deep_ep/buffer.py` (+42 -5) 📝 `tests/test_internode.py` (+2 -2) 📝 `tests/test_low_latency.py` (+11 -1) </details> ### 📄 Description _No description provided_ --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-27 21:05:50 +00:00