deepseek-ai/DeepEP
1
0
Fork 0
mirror of https://github.com/deepseek-ai/DeepEP.git synced 2026-11-05 00:47:32 +00:00
Code Issues 308 Projects Packages Wiki Activity

[PR #580] [MERGED] Refactor, add docu and fix the bug that occurs when dispatch and combine are used consecutively. #645

New issue
Closed
opened 2026-10-27 21:04:49 +00:00 by bsakaguchi · 0 comments
bsakaguchi commented 2026-10-27 21:04:49 +00:00
Owner
Copy link

📋 Pull Request Information

Original PR: https://github.com/deepseek-ai/DeepEP/pull/580
Author: @Autumn1998
Created: 2/5/2026
Status: ✅ Merged
Merged: 2/9/2026
Merged by: @jershi425

Base: hybrid-ep ← Head: refactor

📝 Commits (10+)

📊 Changes

22 files changed (+1258 additions, -662 deletions)

View changed files

📝 csrc/hybrid_ep/allocator/allocator.cu (+1 -0)
📝 csrc/hybrid_ep/backend/hybrid_ep_backend.cuh (+12 -2)
📝 csrc/hybrid_ep/buffer/internode.cu (+71 -8)
📝 csrc/hybrid_ep/buffer/internode.cuh (+42 -7)
➕ csrc/hybrid_ep/buffer/intranode.cu (+336 -0)
➕ csrc/hybrid_ep/buffer/intranode.cuh (+77 -0)
📝 csrc/hybrid_ep/config.cuh (+8 -2)
📝 csrc/hybrid_ep/executor/executor.cu (+95 -59)
📝 csrc/hybrid_ep/executor/executor.cuh (+24 -7)
📝 csrc/hybrid_ep/extension/allgather.cu (+42 -18)
📝 csrc/hybrid_ep/extension/allgather.cuh (+10 -8)
📝 csrc/hybrid_ep/hybrid_ep.cu (+44 -366)
📝 csrc/hybrid_ep/hybrid_ep.cuh (+3 -31)
📝 csrc/hybrid_ep/utils.cuh (+0 -55)
📝 deep_ep/hybrid_ep_buffer.py (+2 -0)
➕ docs/Hybrid-EP_Implementation.md (+471 -0)
📝 docs/README_Hybrid-EP.md (+9 -98)
➕ figures/hybrid-ep-img/Hybrid-EP-workflow.svg (+4 -0)
➕ figures/hybrid-ep-img/RegisterBuffer.png (+0 -0)
➕ figures/hybrid-ep-img/hybrid_ep_combine.png (+0 -0)

...and 2 more files

📄 Description

  1. Refactor
  2. There is a race condition when we use dispatch -> dispatch -> combine -> combine
  3. Add detailed design doc for hybrid-ep

🔄 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/580 **Author:** [@Autumn1998](https://github.com/Autumn1998) **Created:** 2/5/2026 **Status:** ✅ Merged **Merged:** 2/9/2026 **Merged by:** [@jershi425](https://github.com/jershi425) **Base:** `hybrid-ep` ← **Head:** `refactor` --- ### 📝 Commits (10+) - [`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 - [`c07954d`](https://github.com/deepseek-ai/DeepEP/commit/c07954da3c7a1abaa858b8f3fb44ec019b5c2e4e) update parameter - [`37a54cc`](https://github.com/deepseek-ai/DeepEP/commit/37a54ccc25ef260e2d89f25ffa76efee2169b194) add barrier - [`f7dd2f0`](https://github.com/deepseek-ai/DeepEP/commit/f7dd2f0343d9614a9498300ccbea16aff59cd435) add input check - [`2540b8a`](https://github.com/deepseek-ai/DeepEP/commit/2540b8abfc7551597b259c05132ef1d995f0885f) refactor init - [`3122d6e`](https://github.com/deepseek-ai/DeepEP/commit/3122d6eafdfeae7f139f2822da8154dd22a0f990) finish refatore - [`4b5cbd7`](https://github.com/deepseek-ai/DeepEP/commit/4b5cbd7aa90b04d3cf15834b67596f0de7236733) fix RDMA - [`07eea0c`](https://github.com/deepseek-ai/DeepEP/commit/07eea0c0e818d66e28a8dc045ead6dc9c73e17d2) refine doc - [`f942cb2`](https://github.com/deepseek-ai/DeepEP/commit/f942cb29ab82c49b0976d42be02beb547d879355) refine docs ### 📊 Changes **22 files changed** (+1258 additions, -662 deletions) <details> <summary>View changed files</summary> 📝 `csrc/hybrid_ep/allocator/allocator.cu` (+1 -0) 📝 `csrc/hybrid_ep/backend/hybrid_ep_backend.cuh` (+12 -2) 📝 `csrc/hybrid_ep/buffer/internode.cu` (+71 -8) 📝 `csrc/hybrid_ep/buffer/internode.cuh` (+42 -7) ➕ `csrc/hybrid_ep/buffer/intranode.cu` (+336 -0) ➕ `csrc/hybrid_ep/buffer/intranode.cuh` (+77 -0) 📝 `csrc/hybrid_ep/config.cuh` (+8 -2) 📝 `csrc/hybrid_ep/executor/executor.cu` (+95 -59) 📝 `csrc/hybrid_ep/executor/executor.cuh` (+24 -7) 📝 `csrc/hybrid_ep/extension/allgather.cu` (+42 -18) 📝 `csrc/hybrid_ep/extension/allgather.cuh` (+10 -8) 📝 `csrc/hybrid_ep/hybrid_ep.cu` (+44 -366) 📝 `csrc/hybrid_ep/hybrid_ep.cuh` (+3 -31) 📝 `csrc/hybrid_ep/utils.cuh` (+0 -55) 📝 `deep_ep/hybrid_ep_buffer.py` (+2 -0) ➕ `docs/Hybrid-EP_Implementation.md` (+471 -0) 📝 `docs/README_Hybrid-EP.md` (+9 -98) ➕ `figures/hybrid-ep-img/Hybrid-EP-workflow.svg` (+4 -0) ➕ `figures/hybrid-ep-img/RegisterBuffer.png` (+0 -0) ➕ `figures/hybrid-ep-img/hybrid_ep_combine.png` (+0 -0) _...and 2 more files_ </details> ### 📄 Description 1. Refactor 2. There is a race condition when we use dispatch -> dispatch -> combine -> combine 3. Add detailed design doc for hybrid-ep --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
bsakaguchi 2026-10-27 21:04:49 +00:00
Sign in to join this conversation.
No Branch/Tag specified
Branches Tags
main
hybrid-ep
chore/remove-clang-format-yapf
v2.5
nvDev
try-fix-cu128
mori-ep
antgroup-opt
compat-old-torch
add-linter
eager-rdma
v1.2.1
Labels
Clear labels   
pull-request
Mirrored from GitHub Pull Request

No labels
pull-request
Milestone
Clear milestone
No items
No milestone
Projects
Clear projects
No items
No project
Assignees
Clear assignees
No assignees
1 participant
Notifications
Due date
The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set.

Reference
deepseek-ai/DeepEP#645
No description provided.