[PR #6] [MERGED] perf(fsdp2): use KT BF16 reduction for rank-local gradients #6

Closed
opened 2026-10-29 14:17:58 +00:00 by bsakaguchi · 0 comments
Owner

📋 Pull Request Information

Original PR: https://github.com/kvcache-ai/accelerate/pull/6
Author: @yyj6666667
Created: 10/10/2026
Status: ✅ Merged
Merged: 10/10/2026
Merged by: @yyj6666667

Base: main ← Head: perf/fsdp2-bf16-grad-norm-20261010


📝 Commits (1)

  • b4b4260 perf(fsdp2): use KT BF16 reduction for rank-local gradients

📊 Changes

1 file changed (+22 additions, -1 deletions)

View changed files

📝 src/accelerate/accelerator.py (+22 -1)

📄 Description

调用KT侧实现的融合算子


🔄 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/kvcache-ai/accelerate/pull/6 **Author:** [@yyj6666667](https://github.com/yyj6666667) **Created:** 10/10/2026 **Status:** ✅ Merged **Merged:** 10/10/2026 **Merged by:** [@yyj6666667](https://github.com/yyj6666667) **Base:** `main` ← **Head:** `perf/fsdp2-bf16-grad-norm-20261010` --- ### 📝 Commits (1) - [`b4b4260`](https://github.com/kvcache-ai/accelerate/commit/b4b42607164f5a50d875cb8238cd16300c70780d) perf(fsdp2): use KT BF16 reduction for rank-local gradients ### 📊 Changes **1 file changed** (+22 additions, -1 deletions) <details> <summary>View changed files</summary> 📝 `src/accelerate/accelerator.py` (+22 -1) </details> ### 📄 Description 调用KT侧实现的融合算子 --- <sub>🔄 This issue represents a GitHub Pull Request. It cannot be merged through Gitea due to API limitations.</sub>
bsakaguchi 2026-10-29 14:17:58 +00:00
Sign in to join this conversation.
No labels
pull-request
No milestone
No project
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
kvcache-ai/accelerate#6
No description provided.