Performance Optimization: Optimized TileShape Configuration for f8#3617

Closed

MatrixAssembler wants to merge 3 commits intopytorch:mainfrom

MatrixAssembler:opt/tileshape-f8-add-cooperative

Commits on Jan 27, 2025

perf: optimize TileShape configuration for f8
MatrixAssembler
committed

Commits on Jan 28, 2025

style: fix code formatting
MatrixAssembler
committed

Commits on Feb 25, 2025

Reverting f8f8bf16_rowwise.cu to its original state; modifications will be introduced in a future PR with a more efficente selection heuristic approach.
MatrixAssembler
committed