Commits

Branch selector

 main

User selector

drisspg

Datepicker

All time

Commit History

Commits on Mar 31, 2023

• Fix format bug in NT docs (pytorch#97998)

  Show description for a5b6f10
  

  drisspg

  authored and

  pytorchmergebot

  committed
  a5b6f10

  Copy full SHA for a5b6f10

Commits on Mar 30, 2023

• Add NestedTensor ops: logical_not, logical_not_, masked_fill (pytorch#97934)

  Show description for 5a81508
  

  drisspg

  authored and

  pytorchmergebot

  committed
  5a81508

  Copy full SHA for 5a81508

Commits on Mar 29, 2023

• add various NT ops needed for testing (pytorch#97837)

  Show description for f603873
  

  drisspg

  authored and

  pytorchmergebot

  committed
  f603873

  Copy full SHA for f603873

Commits on Mar 22, 2023

• Update NestedTensor add to support non identical striding for NT+NT (pytorch#97195)

  Show description for db15d19
  

  drisspg

  authored and

  pytorchmergebot

  committed
  db15d19

  Copy full SHA for db15d19

Commits on Mar 20, 2023

• Add forward and backward support for silu to NestedTensors (pytorch#97181)

  Show description for a269e5f
  

  drisspg

  authored and

  pytorchmergebot

  committed
  a269e5f

  Copy full SHA for a269e5f

• Add helpful pretty pretting summaries to torch for lldb debugging (pytorch#97101)

  Show description for dba9487
  

  drisspg

  authored and

  pytorchmergebot

  committed
  dba9487

  Copy full SHA for dba9487

Commits on Mar 19, 2023

• Update FlashAttention to work with sm90 Gpus (pytorch#97051)

  Show description for 90537a7
  

  drisspg

  authored and

  pytorchmergebot

  committed
  90537a7

  Copy full SHA for 90537a7

Commits on Mar 17, 2023

• [SDPA] Remove the chunk_grad from mem-eff attention (pytorch#96880)

  Show description for 98a5cf0
  

  drisspg

  authored and

  pytorchmergebot

  committed
  98a5cf0

  Copy full SHA for 98a5cf0

Commits on Mar 16, 2023

• Updates to the release notes scripts and documentation (pytorch#94560)

  Show description for dcafe3f
  

  drisspg

  authored and

  pytorchmergebot

  committed
  dcafe3f

  Copy full SHA for dcafe3f

Commits on Mar 15, 2023

• Fixes a layer_norm_nested backwards edge case. (pytorch#96788)

  Show description for 5612aa6
  

  drisspg

  authored and

  pytorchmergebot

  committed
  5612aa6

  Copy full SHA for 5612aa6

Commits on Mar 14, 2023

• Moves SDPA backward helper native function to functionsmanual.cpp (pytorch#95821)

  Show description for 7ec0d6f
  

  drisspg

  authored and

  pytorchmergebot

  committed
  7ec0d6f

  Copy full SHA for 7ec0d6f

• Fix calls to sizes to enable dynamic shapes with sdpa (pytorch#96674)

  Show description for f56cb41
  

  drisspg

  authored and

  pytorchmergebot

  committed
  f56cb41

  Copy full SHA for f56cb41

Commits on Mar 13, 2023

• Add torch.nn.LayerNorm() to documented list of supported nested tensor ops (pytorch#96434)

  Show description for f330281
  

  drisspg

  authored and

  pytorchmergebot

  committed
  f330281

  Copy full SHA for f330281

Commits on Mar 8, 2023

• [SDPA] Add an optional scale kwarg (pytorch#95259)

  Show description for 11aab72
  

  drisspg

  authored and

  pytorchmergebot

  committed
  11aab72

  Copy full SHA for 11aab72

Commits on Mar 1, 2023

• Rearrange some transformer tests (pytorch#95745)

  Show description for 3fa9396
  

  drisspg

  authored and

  pytorchmergebot

  committed
  3fa9396

  Copy full SHA for 3fa9396

Commits on Feb 28, 2023

• Enabling FlashAttention for SDPA when given NestedTensor (pytorch#95438)

  Show description for 8061478
  

  drisspg

  authored and

  pytorchmergebot

  committed
  8061478

  Copy full SHA for 8061478

Commits on Feb 23, 2023

• [SDPA] Fix bug in parsing scaled_dot_product_attention arguments (pytorch#95311)

  Show description for 29c235e
  

  drisspg

  authored and

  pytorchmergebot

  committed
  29c235e

  Copy full SHA for 29c235e

Commits on Feb 16, 2023

• [SDPA] Update dispatch logic to check for sm86 and head_size == 128 for flash attention (pytorch#94921)

  Show description for 07bc6b9
  

  drisspg

  authored and

  pytorchmergebot

  committed
  07bc6b9

  Copy full SHA for 07bc6b9

• add backwards for layer norm nested (pytorch#94781)

  Show description for 0d7913c
  

  drisspg

  authored and

  pytorchmergebot

  committed
  0d7913c

  Copy full SHA for 0d7913c

Commits on Feb 15, 2023

• add reproducibility notes to nn.UnpoolND operations (pytorch#94629)

  Show description for 7f7f91e
  

  drisspg

  authored and

  pytorchmergebot

  committed
  7f7f91e

  Copy full SHA for 7f7f91e

Commits on Feb 14, 2023

• add backwards for gelu and relu on nested tensors. (pytorch#94776)

  Show description for 63bf767
  

  drisspg

  authored and

  pytorchmergebot

  committed
  63bf767

  Copy full SHA for 63bf767

Commits on Feb 10, 2023

• [SDPA] update type hint for scaled_dot_product_attention and documentation (pytorch#94008)

  Show description for 70026aa
  

  drisspg

  authored and

  pytorchmergebot

  committed
  70026aa

  Copy full SHA for 70026aa

Commits on Feb 9, 2023

• Add back in registration (pytorch#94452)

  Show description for 760836f
  

  drisspg

  authored and

  pytorchmergebot

  committed
  760836f

  Copy full SHA for 760836f

• [SDPA] Adds basic correctness checks (pytorch#94274)

  Show description for 81bbee7
  

  drisspg

  authored and

  pytorchmergebot

  committed
  81bbee7

  Copy full SHA for 81bbee7

Commits on Feb 7, 2023

• [NestedTensor] Call contiguous in linear backward (pytorch#94317)

  Show description for bef2483
  

  drisspg

  authored and

  pytorchmergebot

  committed
  bef2483

  Copy full SHA for bef2483

• [SDPA] Add expanded autograd testing for fused kernels and disable head_dim128 sm86 mem-efficient (pytorch#94009)

  Show description for e0950fc
  

  drisspg

  authored and

  pytorchmergebot

  committed
  e0950fc

  Copy full SHA for e0950fc

Commits on Feb 3, 2023

• [SDPA] Remove private version and only utilize public version (pytorch#94004)

  Show description for 3df0e26
  

  drisspg

  authored and

  pytorchmergebot

  committed
  3df0e26

  Copy full SHA for 3df0e26

Commits on Feb 2, 2023

• [SDPA] Wire up FlashAttention's backward (pytorch#92917)

  Show description for 653dc73
  

  drisspg

  authored and

  pytorchmergebot

  committed
  653dc73

  Copy full SHA for 653dc73

• Add torch_compile_debug/ to .gitignore (pytorch#93889)

  Show description for 68b06ee
  

  drisspg

  authored and

  pytorchmergebot

  committed
  68b06ee

  Copy full SHA for 68b06ee

Commits on Jan 28, 2023

• Use the old aten underscored function for Predictor (pytorch#93096)

  Show description for ca8f5e1
  

  drisspg

  authored and

  pytorchmergebot

  committed
  ca8f5e1

  Copy full SHA for ca8f5e1

Commits on Jan 25, 2023

• Remove windows check for cmake to build Fused kernels (pytorch#91909)

  Show description for a3715ef
  

  drisspg

  authored and

  pytorchmergebot

  committed
  a3715ef

  Copy full SHA for a3715ef

Commits on Jan 24, 2023

• Add USE_FLASH_ATTENTION flag to setup.py (pytorch#92903)

  Show description for 4bc0491
  

  drisspg

  authored and

  pytorchmergebot

  committed
  4bc0491

  Copy full SHA for 4bc0491

• [SDPA] Integrating the main branch of flash_attn instead of cutlass (pytorch#91994)

  Show description for 9613395
  

  drisspg

  authored and

  pytorchmergebot

  committed
  9613395

  Copy full SHA for 9613395

Commits on Jan 23, 2023

• [SDPA] Update SDPA API and make function Public (pytorch#92189)

  Show description for df14650
  

  drisspg

  authored and

  pytorchmergebot

  committed
  df14650

  Copy full SHA for df14650

Commits on Jan 17, 2023

• Update sdp_utils to check gradmode and subclassed tensors (pytorch#92323)

  Show description for 40d6f2a
  

  drisspg

  authored and

  pytorchmergebot

  committed
  40d6f2a

  Copy full SHA for 40d6f2a

Pagination

Previous

Next