[PR #572] [MERGED] [BUG FIX] fix int32 overflow, add switch for experimental features #640

New issue

Closed

opened 2026-10-27 21:04:47 +00:00 by bsakaguchi · 0 comments

bsakaguchi commented

2026-10-27 21:04:47 +00:00

Owner

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/572
Author: @Autumn1998
Created: 1/27/2026
Status: ✅ Merged
Merged: 1/27/2026
Merged by: @jershi425

Base: hybrid-ep ← Head: WAR

📝 Commits (3)

bcc3813 fix int32 overflow
b00165c add switch for custom ag
15073aa update parameter

📊 Changes

4 files changed (+26 additions, -22 deletions)

View changed files

📝 csrc/hybrid_ep/backend/hybrid_ep_backend.cuh (+15 -15)
📝 csrc/hybrid_ep/executor/executor.cu (+3 -3)
📝 deep_ep/hybrid_ep_buffer.py (+7 -3)
📝 tests/utils.py (+1 -1)

📄 Description

fix the int32 overflow on dispatch/combine kernel
add switch for the experimental features => (1) jit cache (2) The dispatch and combine use the same output/input buffer (3) custom allgather

_{🔄 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/572 **Author:** [@Autumn1998](https://github.com/Autumn1998) **Created:** 1/27/2026 **Status:** ✅ Merged **Merged:** 1/27/2026 **Merged by:** [@jershi425](https://github.com/jershi425) **Base:** `hybrid-ep` ← **Head:** `WAR` --- ### 📝 Commits (3) - [`bcc3813`](https://github.com/deepseek-ai/DeepEP/commit/bcc3813c3ff2586891d35d03c743414685d857a3) fix int32 overflow - [`b00165c`](https://github.com/deepseek-ai/DeepEP/commit/b00165cef7f06d02ebb797dc1a222ac57c28ce9d) add switch for custom ag - [`15073aa`](https://github.com/deepseek-ai/DeepEP/commit/15073aa0fd6fa35769476ecb95a6f6e146f88654) update parameter ### 📊 Changes **4 files changed** (+26 additions, -22 deletions) <details> <summary>View changed files</summary> 📝 `csrc/hybrid_ep/backend/hybrid_ep_backend.cuh` (+15 -15) 📝 `csrc/hybrid_ep/executor/executor.cu` (+3 -3) 📝 `deep_ep/hybrid_ep_buffer.py` (+7 -3) 📝 `tests/utils.py` (+1 -1) </details> ### 📄 Description 1. fix the int32 overflow on dispatch/combine kernel 2. add switch for the experimental features => (1) jit cache (2) The dispatch and combine use the same output/input buffer (3) custom allgather --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>

bsakaguchi

2026-10-27 21:04:47 +00:00