Add DCP compatibility for FSDP2-TP sharding in TransformerEngine.#2713
Draft
cspades wants to merge 2 commits intoNVIDIA:mainfrom
Draft
Add DCP compatibility for FSDP2-TP sharding in TransformerEngine.#2713cspades wants to merge 2 commits intoNVIDIA:mainfrom
cspades wants to merge 2 commits intoNVIDIA:mainfrom