Replace has_warp_spec with HAS_AUTO_WS env variable check#205

Closed

ardaunal wants to merge 18 commits intopytorch-labs:mainfrom

ardaunal:ardau/replace-has-warp-spec-with-has-auto-ws

Commits on Apr 15, 2025

update tma_ws
xuzhao9
committed

Commits on Apr 18, 2025

Replace has_warp_spec with HAS_AUTO_WS env variable check
ardaunal
committed

Commits on Apr 30, 2025

Merge branch 'main' into ardau/replace-has-warp-spec-with-has-auto-ws
ardaunal
authored

Commits on May 2, 2025

Merge remote-tracking branch 'remotes/upstream/xz9/flash_attn_tma_ws_persistent' into ardau/replace-has-warp-spec-with-has-auto-ws
ardaunal
committed

Commits on May 7, 2025

Fix TK benchmark under buck; Include "seq_len_kv" in input shape

y-sq
authored and
ardaunal
committed
Add gemm analyzer and input loaders

xuzhao9
authored and
ardaunal
committed
Fix failed internal tests

xuzhao9
authored and
ardaunal
committed
Remove low_mem_dropout in nightly tflops run. (#216)

xuzhao9
authored and
ardaunal
committed
enabling flash_attention (supported variants) for the dashboards.

Dhruva Kaushal
authored and
ardaunal
committed
turning on prefill (llama) shapes for fp8_gemm_rowwise b200 dashboard

Dhruva Kaushal
authored and
ardaunal
committed
pytorch/tritonbench

generatedunixname89002005287564
authored and
ardaunal
committed
Only load input loader in fbcode (#217)

xuzhao9
authored and
ardaunal
committed
Enable ThunderKittens in unit test (#218)

xuzhao9
authored and
ardaunal
committed
Match two autotune configs

xuzhao9
authored and
ardaunal
committed
Load operator to kernel mapping from metadata

xuzhao9
authored and
ardaunal
committed
Add input loader for jagged_dense_dense_sum

xuzhao9
authored and
ardaunal
committed
update tma_ws

xuzhao9
authored and
ardaunal
committed

Commits on May 9, 2025

Merge branch 'pytorch-labs:main' into ardau/replace-has-warp-spec-with-has-auto-ws
ardaunal
authored