Enable MultiCastTranspose for expert weights#628

Open

sudhu2k wants to merge 7 commits into

devfrom

sudhu/grouped_linear_mct

Commits on Jun 10, 2026

Add get_multi_weight_workspace method for optimized weight handling
sudhu2k
committed

Commits on Jun 11, 2026

Refactor quantization logic in TransformerEngineBaseModule for improved workspace management
sudhu2k
committed
Enhance multi_tensor_quantize function to support output tensors
sudhu2k
committed

Commits on Jun 23, 2026

Merge remote-tracking branch 'origin/dev' into sudhu/grouped_linear_mct
sudhu2k
committed
Add MXFP8 support to get_multi_weight_workspace which uses the new multi-tensor quantize kernel
sudhu2k
committed

Commits on Jun 26, 2026

Add comment
sudhu2k
committed

Commits on Jul 1, 2026

Merge remote-tracking branch 'origin/dev' into sudhu/grouped_linear_mct
sudhu2k
committed