Commit Graph

  • b3861ac8e7 [reland] Warn if AccumulateGrad stream does not match producer node stream (#166136) master soulitzer 2025-11-01 12:33:48 +0000
  • 4cc64d6234 [inductor] pre grad graph bisecting (#166344) Shunting Zhang 2025-10-31 17:43:55 -0700
  • 1aef88c72d Avoid DDE in narrow with unbacked start (#166361) Laith Sakka 2025-10-29 15:27:14 -0700
  • f0745ddb11 Replace c10::call_once with static initialization (#166381) Yuanyuan Chen 2025-11-01 07:09:40 +0000
  • 4316df857c [3.14] Fix torch.package.importer (#166767) Nikita Shulga 2025-10-31 16:28:18 -0700
  • 9d6597b1e9 Correctly use test parameters (#166726) Yuanyuan Chen 2025-11-01 04:43:31 +0000
  • e8fadba28c [pytree] add treespec_{leaf,tuple,dict} functions for args_spec modification (#160843) Xuehai Pan 2025-11-01 08:48:26 +0800
  • 60333de85d Revert "Remove setup-env instructions; it's confusing (#166749)" PyTorch MergeBot 2025-11-01 02:55:56 +0000
  • 3dc92d69ed Remove setup-env instructions; it's confusing (#166749) Edward Yang 2025-10-31 21:45:30 -0400
  • f91899ca6c [2/N] Add strict parameter to Python zip calls (#166257) Yuanyuan Chen 2025-11-01 00:35:41 +0000
  • e2dc32f4ba Replace decltype(auto) with auto (#166537) Yuanyuan Chen 2025-11-01 00:30:23 +0000
  • 83cc38d9c1 [precompile] Preserve default arguments for dynamo capture (#166654) Zhengxu Chen 2025-11-01 00:12:10 +0000
  • 8d599045cf add shape check for avg_pool2d (#161952) Sun, Jiayi 2025-10-30 17:34:43 +0000
  • fd5da81fdd [AI Codemod][DevmateFBSourceTestFailureBot] Fix for T243177299 ("Your diff, D85182174, broke some tests") (#166753) Paul de Supinski 2025-10-31 22:49:56 +0000
  • 9261a1fb12 [MPS] Error out when BatchNorm is called for Complex (#166215) Nikita Shulga 2025-10-31 14:07:24 -0700
  • d80ae738c9 compile_worker: Make a timer class (#166465) clr 2025-10-31 13:24:00 -0700
  • 51667435f5 [FlexFlash] Wire up mask_mod + blockmask to flash impl (#166359) drisspg 2025-10-31 18:05:10 +0000
  • 2699f5410b Revert "[xpu][feature] Integrate OneDNN SDPA training forward/backward into XPU OVERRIDEABLE Backend (#162454)" PyTorch MergeBot 2025-10-31 21:58:52 +0000
  • 9970fb97ff Fix Tril Triu SymInt (#166627) Parshant Sharma 2025-10-31 21:53:16 +0000
  • dfebdcab86 [GraphPartition] cache get_free_symbol_uses (#166338) Boyuan Feng 2025-10-31 21:24:05 +0000
  • b09fb481e0 [CD] Upgrade GCC version to 13 for XPU build (#162474) Wang, Chuanqi 2025-10-31 21:15:32 +0000
  • 4e7232c5da [MPS] Fix smooth_l1_loss backward for fp16 (#166687) Nikita Shulga 2025-10-31 14:07:23 -0700
  • 93a70c717a Revert "Add CUDA MXFP4 scaled mm support via. FBGEMM (#166526)" PyTorch MergeBot 2025-10-31 21:10:28 +0000
  • d97144d31e [5/N] Remove unused loop variables in tests (#166716) Yuanyuan Chen 2025-10-31 20:47:54 +0000
  • e4043884c7 [dynamo, 3.14] fix segfault due to improper create_call_function_ex (#166678) William Wen 2025-10-30 17:25:01 -0700
  • 4a7bc1d522 [BE][Typing][Dynamo] Type misc files in torch/_dynamo/variables/ (#166569) Lucas Kabela 2025-10-31 20:42:23 +0000
  • 8209a0506b [Pytorch] Enable aarch64 convert autovec only on clang (#166739) Nicolas De Carli 2025-10-31 20:22:33 +0000
  • 70aeb49198 [dynamo] clarify graph break handling/logging in symbolic_convert (#166587) William Wen 2025-10-31 03:51:00 +0000
  • cf9a834f39 [BE] Move GreenContext implementation details to cpp (#166462) Nikita Shulga 2025-10-31 20:10:59 +0000
  • 856a7a5298 Add missing device to namedtensor tests (#166717) Yuanyuan Chen 2025-10-31 20:04:38 +0000
  • ef8d97efcf fix broken nn_convolution test (#166666) Camyll Harajli 2025-10-31 19:59:50 +0000
  • d2be06f673 [cpu][fix] Update ACL version to fix crashes with tensor sizes > 2^31-1 (#165904) Fadi Arafeh 2025-10-31 16:52:39 +0000
  • 08f4535378 Refactor AOTAutogradCacheEntry into AOTAutogradResult (#166656) James Wu 2025-10-30 22:12:06 -0700
  • 30157d30f0 Add regional aot eager support to AOTAutogradCacheEntry (#166650) James Wu 2025-10-30 09:53:41 -0700
  • b470e59c38 partitioner option to ignore partitioner_tag for abstract usage (#166725) IvanKobzarev 2025-10-31 07:25:06 -0700
  • 85b85f6c2c Revert "[pytree] add treespec_{leaf,tuple,dict} functions for args_spec modification (#160843)" PyTorch MergeBot 2025-10-31 18:31:32 +0000
  • b71966f67b [PyTorch] Improve aarch64 performance of bfloat16 ops - retry (#166028) (#166641) Nicolas De Carli 2025-10-31 18:21:04 +0000
  • 0947765eb9 Cache even more work for return_and_correct_aliasing (#166365) Scott Wolchok 2025-10-30 09:34:35 -0700
  • 239e7b541a [ROCm][CI] upgrade nightly wheels to ROCm 7.1 (#166730) Jeff Daily 2025-10-31 17:30:47 +0000
  • ffaa6578b7 Revise deprecation warning for ONNX exporter (#166692) Justin Chu 2025-10-31 17:23:55 +0000
  • 365ed62f61 Document LibTorch ABI more, add README to headeronly (#166661) Jane Xu 2025-10-30 13:54:13 -0700
  • fcc1063566 Revert "[BE][Typing][Dynamo] Type misc files in torch/_dynamo/variables/ (#166569)" PyTorch MergeBot 2025-10-31 16:59:32 +0000
  • 121235956b update Node.is_impure check if subgraph contains impure ops (#166609) Jazlyn Li 2025-10-31 16:58:15 +0000
  • aa9c96af04 [BE][Typing][Dynamo] Type misc files in torch/_dynamo/variables/ (#166569) Lucas Kabela 2025-10-31 16:56:50 +0000
  • c3b71d5499 [ROCm][CI] remove relaxed tolerance for tf32 tests (#166478) Jeff Daily 2025-10-31 16:15:39 +0000
  • 1e3600b528 [MPS] Move logaddexp/logaddexp2 to Metal and support complex (#166670) Kurt Mohler 2025-10-30 17:52:53 -0500
  • fee7624bd6 [PT2] set choice handler in config (#166607) Xuan Zhang 2025-10-31 15:40:05 +0000
  • 24e94e021a [ROCm][CI] create ROCm 7.1 magma tarball (#166693) Jeff Daily 2025-10-31 15:20:00 +0000
  • 69be99ee51 Remove manually synced arch versions in tools/nightly.py (#166616) Xuehai Pan 2025-10-30 14:47:52 +0800
  • 034e951b0c [CUDA][cuBLASLt] addmm -- extend bias fusions to cases with (1 by n) shapes (#166307) Nikita Vedeneev 2025-10-31 10:18:28 +0000
  • 160ab53dd5 Update weight tensor initialization in RMSNormalization (#166550) Justin Chu 2025-10-31 14:29:27 +0000
  • 5bcfdae71d Revert "Make PT2 compile backprop through custom op without autograd key a hard error (#166367)" PyTorch MergeBot 2025-10-31 13:44:05 +0000
  • 4e8ba37ce3 Revert "[BE] Move GreenContext implementation details to cpp (#166462)" PyTorch MergeBot 2025-10-31 13:20:48 +0000
  • 26534e9809 Revert "[GraphPartition] cache get_free_symbol_uses (#166338)" PyTorch MergeBot 2025-10-31 12:57:54 +0000
  • 657f8c3e21 Revert "Fix torch.full with dynamic tensor fill_value in torch.compile (#166554)" PyTorch MergeBot 2025-10-31 12:55:31 +0000
  • b0831930ed [inductor] Mark / restrict tests that only work if ATen is used for matmul (#166518) Mwiza Kunda 2025-10-31 12:29:03 +0000
  • c01636e1bc Fixes the sparse tensor issue (#163535) arkadip-maitra 2025-10-31 11:48:26 +0000
  • fd68d409ad [xpu][feature] Integrate OneDNN SDPA training forward/backward into XPU OVERRIDEABLE Backend (#162454) fengqing.lu 2025-10-31 11:20:38 +0000
  • 0d3a4f7155 [CD] Enable Inductor performance test for xpu (#166289) Wang, Chuanqi 2025-10-31 10:52:07 +0000
  • 108bb224f7 [pytree] add treespec_{leaf,tuple,dict} functions for args_spec modification (#160843) Xuehai Pan 2025-10-31 14:31:20 +0800
  • fc8ac1216c [4/N] Remove unused loop variables in tests (#166690) Yuanyuan Chen 2025-10-31 10:20:48 +0000
  • 030de07aff [2/N] Use 'is' in callable comparisons (#166685) Yuanyuan Chen 2025-10-31 08:08:07 +0000
  • 7d67a41db4 make FXConverter.generate use V.fake_mode instead of _detect_fake_mode_from_gm (#166591) Jazlyn Li 2025-10-31 05:52:07 +0000
  • 85b035ca9c [nativert] Downcast triton double arguments to floats (#166620) Minjang Kim 2025-10-31 03:52:20 +0000
  • 267d0197bf [dynamo] fix error_on_graph_break bug where non-empty checkpoint results in unwanted graph break resumption (#166586) William Wen 2025-10-31 00:50:46 +0000
  • 1dec8a67a8 [dynamo, nested graph breaks] add disable_nested_graph_breaks decorator/context manager (#166477) William Wen 2025-10-31 00:50:46 +0000
  • 797cd80b26 [dynamo, nested graph breaks] codegen dead nested cells correctly (#166476) William Wen 2025-10-31 00:50:46 +0000
  • 7d39401fa0 Revert "[BE][Typing][Dynamo] Type misc files in torch/_dynamo/variables/ (#166569)" PyTorch MergeBot 2025-10-31 03:31:01 +0000
  • e3ae0594d1 Add CUDA MXFP4 scaled mm support via. FBGEMM (#166526) Simon Layton 2025-10-30 07:46:40 -0700
  • f1e4c42b6e [BE][Typing][Dynamo] Type misc files in torch/_dynamo/variables/ (#166569) Lucas Kabela 2025-10-31 02:57:55 +0000
  • d3e511f07c [Inductor] support masked vectorization for the tail_loop for fp8 datatype (#163324) Sun, Jiayi 2025-10-30 04:36:44 +0000
  • d3be06cbdc [MTIAGraph][Pytorch][2/n] Add binding for Python to C++, and hook for Pytorch to Fbcode (#165963) Andy (An) Wang 2025-10-31 02:52:48 +0000
  • 1129605415 [ROCm][CI] create ROCm 7.1 images for binary builds (#166665) Jeff Daily 2025-10-31 02:52:37 +0000
  • a6b1ef1717 [GraphPartition] cache get_free_symbol_uses (#166338) Boyuan Feng 2025-10-31 02:50:10 +0000
  • 12577064dd [MPS] Fix crash when max/min ops called for complex types (#166214) Nikita Shulga 2025-10-27 18:07:20 -0700
  • 24b6eb7727 [Inductor] Enable Custom op Autotune Decompositions and Parameter Tuning (#164212) Tianren Gao 2025-10-31 02:27:57 +0000
  • 32066772b3 Fix torch.full with dynamic tensor fill_value in torch.compile (#166554) Amal Dev Haridevan 2025-10-31 00:55:58 +0000
  • 47f0024310 [CI][BE] Factor out repeated test code (#166481) Nikita Shulga 2025-10-30 14:28:04 -0700
  • 98d640bb11 Remove AT_USE_HIPSPARSE_GENERIC_API (#166393) Yuanyuan Chen 2025-10-31 00:49:07 +0000
  • 5d288bc3f7 [BE] Move GreenContext implementation details to cpp (#166462) Nikita Shulga 2025-10-31 00:48:01 +0000
  • bfb47ec50e [dynamo] support tracing new typing union syntax X | Y (#166599) William Wen 2025-10-30 11:23:24 -0700
  • 7a0cd8ed09 [ROCm] Disable __builtin_amdgcn_rcpf for gfx90a (#166454) Prachi Gupta 2025-10-30 23:38:55 +0000
  • 984e64b2cd [inductor] Fix constant folder (#166655) angelayi 2025-10-30 22:51:25 +0000
  • b9bcb37f40 [DebugMode] store stringify args by default (#166347) Pian Pawakapan 2025-10-28 17:57:22 -0700
  • 7e3b9d105e [CP][BE][2/2] Refactor the code structure (#166501) Chien-Chin Huang 2025-10-30 09:43:01 -0700
  • 45c3f02d69 [ROCm] moved gfx1100 back to experimental status for AOTriton (#166397) Artem Kuzmitckii 2025-10-30 21:43:01 +0000
  • f5543e3741 [wip] fix searchsorted non dense (#165064) eellison 2025-10-30 09:11:31 -0700
  • 5fc2c7a2a1 [ROCm][inductor] More configs for pointwise kernels. (#166470) Nichols A. Romero 2025-10-30 21:20:08 +0000
  • 7692fa09cd [Code Clean] Clean asserts in torch/ao/quantization/fx/* (#165420) zhudada 2025-10-30 20:53:31 +0000
  • df71b70727 [cuDNN][conv] Re-enable cuDNN for 3D convolutions (fixed in 9.15+) (#166480) Eddie Yan 2025-10-30 20:47:20 +0000
  • 80ba6e458f Add warning when users have incomplete setup for type checking (#166603) Maggie Moss 2025-10-30 20:37:40 +0000
  • 0d50e5d8d4 [3/N] Fix unused loop variables (#166509) Yuanyuan Chen 2025-10-30 20:13:47 +0000
  • 99b05d1b78 Better 1x128, 128x128 error handling on non-Hopper (#166639) Simon Layton 2025-10-30 08:07:46 -0700
  • f911d64750 [CUDA] xFail max-autotune grouped gemm tests on devices with insufficient SM count (#165921) Eddie Yan 2025-10-30 20:05:03 +0000
  • 52db60170d Enable verify_dynamo on Python 3.13 (#166497) Yuanyuan Chen 2025-10-30 19:52:28 +0000
  • 56838bad5f [CP][BE][1/2] Refactor the code structure (#166456) Chien-Chin Huang 2025-10-30 09:43:01 -0700
  • ad3a56ab98 Add a compile-time flag to trigger verbose logging for device-side asserts (#166171) Sudarshan Raghunathan 2025-10-30 19:43:44 +0000
  • a7fd0b4001 [ROCm][CI] fix disk space message (#166645) amdfaa 2025-10-30 19:38:31 +0000
  • 181ee3bd42 fix: Add missing signals_to_handle to launcher logging (#166631) leopold-tzafon 2025-10-30 19:31:21 +0000
  • 0ec0549823 Introduce a new API torch.xpu.get_per_process_memory_fraction (#165511) Yu, Guangye 2025-10-15 23:38:06 +0000