DeepSpeed
[BUG] Fix: Fix gradient norm calculation and dynamic shape blocking in PP+ZeRO1 collective communication
#7847
Open

[BUG] Fix: Fix gradient norm calculation and dynamic shape blocking in PP+ZeRO1 collective communication #7847

Thinksky5124 wants to merge 1 commit into deepspeedai:master from Thinksky5124:master
Thinksky5124
Thinksky5124 Fix pp+zero1 bugs
99697687
Thinksky5124 Thinksky5124 requested a review from tjruwase tjruwase 2 days ago
Thinksky5124 Thinksky5124 requested a review from tohtana tohtana 2 days ago
Thinksky5124 Thinksky5124 requested a review from loadams loadams 2 days ago
chatgpt-codex-connector
chatgpt-codex-connector commented on 2026-02-12

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone