Wing Lian b0c0ba7b4d FSDP grad accum fix (#34645) ...

* add gradient accumulation steps tests for fsdp

* invert no_sync context to fix training for fsdp

2024-11-15 22:28:06 +01:00

..

test_fsdp.py

FSDP grad accum fix (#34645)

2024-11-15 22:28:06 +01:00