)]}' { "log": [ { "commit": "ab2ddb1f6c6d6d010f151ca6b4650857225028a5", "tree": "884380c3cb683e4e2bac628a543670d6122f756d", "parents": [ "0f68a9282634efea439d79a814535b715891ea1a", "9ea4d194516179402e5a53f11fce30b4de511cff" ], "author": { "name": "Xin Li", "email": "delphij@google.com", "time": "Fri Dec 20 13:39:41 2024 -0800" }, "committer": { "name": "Xin Li", "email": "delphij@google.com", "time": "Fri Dec 20 13:39:41 2024 -0800" }, "message": "Empty merge ab/12770256 into aosp-main-future\n\nMerged-In: Ie8e06384881420f5c89f761a86d6bf8dfeb35c57\nChange-Id: Ic89411722fd6f49668d20efb14032c9fcb68771b\n" }, { "commit": "0f68a9282634efea439d79a814535b715891ea1a", "tree": "884380c3cb683e4e2bac628a543670d6122f756d", "parents": [ "db6a337ffc1f810148b980b8de4a0ed8449ca896", "9b654ccb26cdaa2a75ba1d1dab4778ebf67f571d" ], "author": { "name": "Sadaf Ebrahimi", "email": "sadafebrahimi@google.com", "time": "Thu Dec 19 08:09:18 2024 -0800" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Thu Dec 19 08:09:18 2024 -0800" }, "message": "Add janitors to the OWNERS file am: 9b654ccb26\n\nOriginal change: https://android-review.googlesource.com/c/platform/external/pytorch/+/3423989\n\nChange-Id: Icc0559bdb55c8939417167cf387a6d57594938c3\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "9b654ccb26cdaa2a75ba1d1dab4778ebf67f571d", "tree": "884380c3cb683e4e2bac628a543670d6122f756d", "parents": [ "db6a337ffc1f810148b980b8de4a0ed8449ca896" ], "author": { "name": "Sadaf Ebrahimi", "email": "sadafebrahimi@google.com", "time": "Wed Dec 18 20:36:27 2024 +0000" }, "committer": { "name": "Sadaf Ebrahimi", "email": "sadafebrahimi@google.com", "time": "Wed Dec 18 20:36:27 2024 +0000" }, "message": "Add janitors to the OWNERS file\n\nTest: TreeHugger\nChange-Id: Ibb8052bded81e7014506c66a3ef9b1cac33b7a26\n" }, { "commit": "9ea4d194516179402e5a53f11fce30b4de511cff", "tree": "4d37b398617f5f1599c7adecf7d118d387569903", "parents": [ "fc756f55f4ca27f6e8b8516a4f1201ba34ee6bc9", "db6a337ffc1f810148b980b8de4a0ed8449ca896" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Thu Oct 31 18:14:06 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Thu Oct 31 18:14:06 2024 +0000" }, "message": "Upgrade pytorch to v2.5.1 am: f6beb24584 am: db6a337ffc\n\nOriginal change: undetermined\n\nChange-Id: Ie6746978b50779c2c7d0ce1627d1cfda00d4cf29\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "db6a337ffc1f810148b980b8de4a0ed8449ca896", "tree": "4d37b398617f5f1599c7adecf7d118d387569903", "parents": [ "5f9952124cae3c98a157efd253450af388a83c9b", "f6beb245846b48d557b1eb5601d4f73b5b47f96b" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Thu Oct 31 17:56:58 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Thu Oct 31 17:56:58 2024 +0000" }, "message": "Upgrade pytorch to v2.5.1 am: f6beb24584\n\nOriginal change: https://android-review.googlesource.com/c/platform/external/pytorch/+/3327513\n\nChange-Id: I953b508e773d107b1f095f4b2ee6429256b4805c\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "f6beb245846b48d557b1eb5601d4f73b5b47f96b", "tree": "4d37b398617f5f1599c7adecf7d118d387569903", "parents": [ "b5ea807cbd9388a4bf5953e5505c4b1681a99138", "a8d6afb511a69687bbb2b7e88a3cf67917e1697e" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Thu Oct 31 00:06:12 2024 +0000" }, "committer": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Thu Oct 31 00:20:55 2024 +0000" }, "message": "Upgrade pytorch to v2.5.1\n\nThis project was upgraded with external_updater.\nUsage: tools/external_updater/updater.sh update external/pytorch\nFor more info, check https://cs.android.com/android/platform/superproject/main/+/main:tools/external_updater/README.md\n\nTest: TreeHugger\nChange-Id: Id0019130e86d74f58ccf5b10972002bccec1c8ff\n" }, { "commit": "fc756f55f4ca27f6e8b8516a4f1201ba34ee6bc9", "tree": "1b6783a8ea7345c52add9c116147906047fce44c", "parents": [ "765c7867ec8991527c62d9029a8b7339ffb5fa95", "5f9952124cae3c98a157efd253450af388a83c9b" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Tue Oct 29 17:38:16 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Tue Oct 29 17:38:16 2024 +0000" }, "message": "Python soong target not support \"apex_available\" am: b5ea807cbd am: 5f9952124c\n\nOriginal change: https://android-review.googlesource.com/c/platform/external/pytorch/+/3323735\n\nChange-Id: Icb50b1043de6258d08a6c63ef006de6766a89264\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "765c7867ec8991527c62d9029a8b7339ffb5fa95", "tree": "ea32d2073d28263fa77b915ed5758ae59465d25d", "parents": [ "9a0dfa64f5ce7ee734f52599353fe32bcb3afa7e", "3f809a3fcfdac6ca1aa68d64bc1f6ad9e6aa7e00" ], "author": { "name": "Inna Palant", "email": "ipalant@google.com", "time": "Tue Oct 29 17:37:55 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Tue Oct 29 17:37:55 2024 +0000" }, "message": "Merge remote-tracking branch \u0027origin/upstream\u0027 am: 6fa82c6c96 am: 3f809a3fcf\n\nOriginal change: https://googleplex-android-review.googlesource.com/c/platform/external/pytorch/+/29954929\n\nChange-Id: I1c8e362407e2160b18e066dc012ae233c521dc71\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "5f9952124cae3c98a157efd253450af388a83c9b", "tree": "1b6783a8ea7345c52add9c116147906047fce44c", "parents": [ "3f809a3fcfdac6ca1aa68d64bc1f6ad9e6aa7e00", "b5ea807cbd9388a4bf5953e5505c4b1681a99138" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Tue Oct 29 16:44:37 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Tue Oct 29 16:44:37 2024 +0000" }, "message": "Python soong target not support \"apex_available\" am: b5ea807cbd\n\nOriginal change: https://android-review.googlesource.com/c/platform/external/pytorch/+/3323735\n\nChange-Id: I5fa01242bbab82c9309ffc7119b13dcbdb3b4b2c\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "3f809a3fcfdac6ca1aa68d64bc1f6ad9e6aa7e00", "tree": "ea32d2073d28263fa77b915ed5758ae59465d25d", "parents": [ "9a0dfa64f5ce7ee734f52599353fe32bcb3afa7e", "6fa82c6c969285e30294f7e4dd4d213ac0ed8ab4" ], "author": { "name": "Inna Palant", "email": "ipalant@google.com", "time": "Tue Oct 29 16:43:52 2024 +0000" }, "committer": { "name": "Automerger Merge Worker", "email": "android-build-automerger-merge-worker@system.gserviceaccount.com", "time": "Tue Oct 29 16:43:52 2024 +0000" }, "message": "Merge remote-tracking branch \u0027origin/upstream\u0027 am: 6fa82c6c96\n\nOriginal change: undetermined\n\nChange-Id: I285197e664adb7ac84c14cd53e2e626748620ea7\nSigned-off-by: Automerger Merge Worker \u003candroid-build-automerger-merge-worker@system.gserviceaccount.com\u003e\n" }, { "commit": "b5ea807cbd9388a4bf5953e5505c4b1681a99138", "tree": "1b6783a8ea7345c52add9c116147906047fce44c", "parents": [ "6fa82c6c969285e30294f7e4dd4d213ac0ed8ab4" ], "author": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Mon Oct 28 22:22:22 2024 +0000" }, "committer": { "name": "Qiao Li", "email": "qiaoli@google.com", "time": "Mon Oct 28 22:25:25 2024 +0000" }, "message": "Python soong target not support \"apex_available\"\n\nChange-Id: I2812bd3109c6aed69910b74292c5d3c7c2ca2cd7\n" }, { "commit": "6fa82c6c969285e30294f7e4dd4d213ac0ed8ab4", "tree": "ea32d2073d28263fa77b915ed5758ae59465d25d", "parents": [ "9a0dfa64f5ce7ee734f52599353fe32bcb3afa7e", "e3a10289adfb8e9f7155218712c99d109f622fc1" ], "author": { "name": "Inna Palant", "email": "ipalant@google.com", "time": "Mon Oct 28 20:20:02 2024 +0000" }, "committer": { "name": "Inna Palant", "email": "ipalant@google.com", "time": "Mon Oct 28 20:20:02 2024 +0000" }, "message": "Merge remote-tracking branch \u0027origin/upstream\u0027\n" }, { "commit": "a8d6afb511a69687bbb2b7e88a3cf67917e1697e", "tree": "8efc6915def0ae723546d2238c95b08d52840ec1", "parents": [ "f31b8bbc5b8834cdac83d1b92e36a8896ed2e4b9" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 18:14:52 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 18:14:52 2024 -0700" }, "message": "Disabling amp context when invoking compiler (#138659)\n\nDisabling amp context when invoking compiler (#138624)\r\n\r\nFix for https://github.com/pytorch/pytorch/issues/133974\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/138624\r\nApproved by: https://github.com/bdhirsh, https://github.com/drisspg\r\n\r\n(cherry picked from commit 5942b2985000e0c69ec955b6c88dee8b5d7e67fd)\r\n\r\nCo-authored-by: eellison \u003celias.ellison@gmail.com\u003e" }, { "commit": "f31b8bbc5b8834cdac83d1b92e36a8896ed2e4b9", "tree": "acb609d70764e45710b0e6b79810edce3843340d", "parents": [ "848e7ac42a3c121d03bcff6cb53c1afb9b76164e" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 16:25:25 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 16:25:25 2024 -0700" }, "message": "[MPS] Fix sliced cast (#138535)\n\n[MPS] Fix sliced cast (#138314)\r\n\r\nThis fixes internal crash due to the invalid bufer size computation if sliced API is used\r\n\r\nNot sure what was the purpose of\r\n```c++\r\nIntArrayRef baseShape;\r\nif (src.is_view()) {\r\n baseShape \u003d src._base().sizes();\r\n} else {\r\n baseShape \u003d getIMPSAllocator()-\u003egetBufferShape(src.storage().data());\r\n}\r\nint flattenedShaped \u003d 1;\r\nfor (const auto i : c10::irange(baseShape.size())) {\r\n flattenedShaped *\u003d baseShape[i];\r\n}\r\n```\r\nAs flattenShaped could be much easier computed as `[srcBuf\r\nlengh]/src.element_size()`, and even if `srcBuf` is padded it\u0027s a safe thing to do.\r\n\r\nWhen someone allocated buffer to hold say uint8 and that view-casted it\r\nto float16, attempt to compute `baseShape` returned sizes of original\r\ntensor in its data type, rather than size in new dtypes\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/137800\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/138314\r\nApproved by: https://github.com/albanD, https://github.com/DenisVieriu97\r\n\r\n(cherry picked from commit de16159e565e7a08294347e31e97ca08a3468227)\r\n\r\nCo-authored-by: Nikita Shulga \u003cnikita.shulga@gmail.com\u003e" }, { "commit": "848e7ac42a3c121d03bcff6cb53c1afb9b76164e", "tree": "7e8eda579ce0f62fdd6209713182703734d01407", "parents": [ "885c823759be505f5725578e0ee46cfec249ccf3" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 15:51:29 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 15:51:29 2024 -0700" }, "message": "[SDPA-CUDNN] Make CuDNN Attention Opt in (#138587)\n\n[SDPA-CUDNN] Make CuDNN Attention Opt in (#138522)\r\n\r\n# Summary\r\nCurrently we have a `cudnn_order` that says on H100 w/ new enough CuDNN backend (we ship a 9.1 version in OSS) try to run CuDNN attention first. We have already encountered a few bugs with the release of 2.5:\r\n\r\n1. https://github.com/pytorch/pytorch/issues/138529\r\n2. https://github.com/huggingface/diffusers/issues/9704\r\n3. https://github.com/pytorch/pytorch/pull/138354\r\n\r\nIn light of the above we are going to make the CuDNN backend Opt-in by default.\r\n\r\nThis can be done easily with the context manager for choosing backends I.e.:\r\n``` Python\r\nfrom torch.nn.attention import sdpa_kernel, SDPBackend\r\n\r\nwith sdpa_kernel(SDPBackend.CUDNN_ATTENTION):\r\n out \u003d F.scaled_dot_product_attention(q, k, v)\r\n\r\n```\r\n\r\nThis PR puts the CuDNN backend as the lowest precedence in the backend list, meaning that the Math backend will always be chosen unless disabled (which is done via the context manager).\r\n\r\nCc @atalman\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/138522\r\nApproved by: https://github.com/ngimel, https://github.com/eqy, https://github.com/malfet\r\n\r\n(cherry picked from commit 9a9a0abc2818d40d06eda6c0b6fdbc949474f12e)\r\n\r\nCo-authored-by: drisspg \u003cdrisspguessous@gmail.com\u003e" }, { "commit": "885c823759be505f5725578e0ee46cfec249ccf3", "tree": "902924b0087ce78003c75bf5b92d5cee02f27933", "parents": [ "8c3ed97baa03668945064b9f1f38496041de2ec0" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 15:47:37 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 15:47:37 2024 -0700" }, "message": "Update doc copyrights to 2024 (#138650)\n\nUpdate copyrights to 2024 (#138638)\r\n\r\nSpiritual successor of https://github.com/pytorch/pytorch/pull/119413 + CPP docs copyright update as well\r\nFixes https://github.com/pytorch/pytorch/issues/138630\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/138638\r\nApproved by: https://github.com/atalman\r\n\r\n(cherry picked from commit d1be61ce4eb31640d1bdce07c8e6b17d03cbdca6)\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "8c3ed97baa03668945064b9f1f38496041de2ec0", "tree": "b505f1ed978c05a96a5887f80bb1414de58fd8ad", "parents": [ "70cf2bbc0b9ebd30b7604b29123a9920d6fcdab6" ], "author": { "name": "Nikita Shulga", "email": "2453524+malfet@users.noreply.github.com", "time": "Tue Oct 22 15:06:53 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 15:06:53 2024 -0700" }, "message": "Update cpuinfo submodule (#138600)\n\nSpiritual cherry-pick of https://github.com/pytorch/pytorch/pull/138351 that picks https://github.com/pytorch/cpuinfo/pull/258 into the branch\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/138333\r\n\r\n\r\nTest Plan: `python -c \"import torch\"` finishes without any output on the screen" }, { "commit": "e3a10289adfb8e9f7155218712c99d109f622fc1", "tree": "ea32d2073d28263fa77b915ed5758ae59465d25d", "parents": [ "ee1b6804381c57161c477caa380a840a84167676" ], "author": { "name": "Emily Fu", "email": "fumengyao@google.com", "time": "Fri Oct 18 19:05:26 2024 +0000" }, "committer": { "name": "Emily Fu", "email": "fumengyao@google.com", "time": "Tue Oct 22 21:04:21 2024 +0000" }, "message": "Include platform/external/pytorch\n\nThird-Party Import of:\nhttps://github.com/pytorch/pytorch/releases/tag/v2.4.1\nRequest Document: go/android3p-revised\nFor CL reviewers: go/android3p#reviewing-a-cl\nFor Build Team: go/ab-third-party-imports\nBug: http://b/372933149\nOriginal import of the code can be found at: https://googleplex-\nandroid.googlesource.com/platform/external/pytorch/+/refs/heads/third-\nparty-review.\nSecurity Questionnaire: http://b/372933149#comment1\n\nConfirmed pytorch can be built locally\n\nTest: third party review\n\nChange-Id: Ie7a476b5b50c5e224f9aa2d0af1f750a714cfaa1\n" }, { "commit": "70cf2bbc0b9ebd30b7604b29123a9920d6fcdab6", "tree": "431604e5a0a423fbd4e9456170bcf86dc857a3ca", "parents": [ "cde6b382ff39210939516f0d0cc3d270ced0fed0" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 12:32:56 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 12:32:56 2024 -0700" }, "message": "Add link to torch.compile the missing manual in troubleshooting (#137369)\n\nAdd link to torch.compile the missing manual in troubleshooting (#137301)\r\n\r\nFixes #ISSUE_NUMBER\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/137301\r\nApproved by: https://github.com/svekars\r\n\r\nCo-authored-by: Svetlana Karslioglu \u003csvekars@meta.com\u003e\r\n(cherry picked from commit 22e19bd2d70409c2908edf0b0d00abb9209e3aaa)\r\n\r\nCo-authored-by: Michael Lazos \u003cmlazos@meta.com\u003e" }, { "commit": "cde6b382ff39210939516f0d0cc3d270ced0fed0", "tree": "f32947b34ba730d78c9e46605929acc17a4d9f39", "parents": [ "4076a738b08e0118f0b0464fe5f79bf4109c29c5" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Oct 22 10:45:53 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 22 10:45:53 2024 -0700" }, "message": "Don\u0027t try to load cufile (#138539)\n\nDon\u0027t try to load cufile (#138501)\r\n\r\nTrying to loading it caused a big issue with 2.5.0 release - https://github.com/pytorch/pytorch/issues/138324\r\n\r\ncufile is not actually used currently by default, see #133489\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/138501\r\nApproved by: https://github.com/atalman, https://github.com/mikaylagawarecki, https://github.com/malfet\r\n\r\n(cherry picked from commit 012ff2a0aaf81bce25b7eda8c0021f5a784c11a6)\r\n\r\nCo-authored-by: Sergii Dymchenko \u003csdym@meta.com\u003e" }, { "commit": "4076a738b08e0118f0b0464fe5f79bf4109c29c5", "tree": "cf96b1435e79ba13e6d1ac0743372b65a07736b8", "parents": [ "a97c15174b8bb4f4cda95f86b759aabc2fbe25df" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Mon Oct 21 21:11:08 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Oct 21 18:11:08 2024 -0700" }, "message": "[Cherry-Pick] Use cuda 12.4 pytorch_extra_install_requirements as default (#138526)\n\nCherry-Picks https://github.com/pytorch/pytorch/pull/138458\r\nNeed to do it manually due to conflict with generated files.\r\n" }, { "commit": "a97c15174b8bb4f4cda95f86b759aabc2fbe25df", "tree": "f973c45c88613d115389aa82d009445ab9101467", "parents": [ "32f585d9346e316e554c8d9bf7548af9f62141fc" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Fri Oct 18 08:24:39 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Oct 18 08:24:39 2024 -0700" }, "message": "update getting started xpu (#138090)\n\nupdate get start xpu (#137479)\r\n\r\n1. respect the comment from the community, downgrade the \"Beta\" to \"Prototype\" for the first xpu release with wheel\r\n2. add wheels installation of torchaudio \u0026 torchvision for nightly on Windows\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/137479\r\nApproved by: https://github.com/atalman, https://github.com/malfet\r\n\r\n(cherry picked from commit 7ba706c74e352dc752a0afd4b23d250108e2cb3e)\r\n\r\nCo-authored-by: Zheng, Zhaoqiong \u003czhaoqiong.zheng@intel.com\u003e" }, { "commit": "9a0dfa64f5ce7ee734f52599353fe32bcb3afa7e", "tree": "4b825dc642cb6eb9a060e54bf8d69288fbee4904", "parents": [], "author": { "name": "Inna Palant (xWF)", "email": "ipalant@google.com", "time": "Fri Oct 18 05:57:25 2024 +0000" }, "committer": { "name": "Android (Google) Code Review", "email": "android-gerrit@google.com", "time": "Fri Oct 18 05:57:25 2024 +0000" }, "message": "Initial empty repository\n" }, { "commit": "32f585d9346e316e554c8d9bf7548af9f62141fc", "tree": "aff7005db8fda4f96f10248c4faaba57d01b538a", "parents": [ "417a0763a7d69f6ce80719ac89c1d2deeee78163" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Mon Oct 14 13:48:44 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Oct 14 13:48:44 2024 -0400" }, "message": "[Release only] use triton 3.1.x from pypi (#137895)\n\n* [Release only] use triton 3.1.x from pypi\r\n\r\n* [Release only] Disable triton build workflows\r\n\r\n* fix" }, { "commit": "417a0763a7d69f6ce80719ac89c1d2deeee78163", "tree": "31061c03fb784dc89acd2173a89b57fb206f9d09", "parents": [ "119e7344d905f69d9b37734becba2ada12641d0c" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Thu Oct 03 11:29:56 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Oct 03 08:29:56 2024 -0700" }, "message": "[split build] move periodic split builds into own concurrency group (#135510) (#137265)\n\nTo avoid nightly workflows cancelling each other\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135510\r\nApproved by: https://github.com/clee2000, https://github.com/huydhn, https://github.com/malfet\r\n\r\nCo-authored-by: Sahan Paliskara \u003csahanp@meta.com\u003e\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "119e7344d905f69d9b37734becba2ada12641d0c", "tree": "8d813d4fe6a4e9f67bf56a76d28cae8fd763cd64", "parents": [ "783a6a424c30869f4b8ea8c686b4ce4991415e5c" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Wed Oct 02 18:04:47 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 18:04:47 2024 -0700" }, "message": "[RELEASE-ONLY CHANGES] Fix dependency on filesystem on Linux (#137242)\n\n* Apply changes from https://github.com/pytorch/pytorch/pull/135374\r\n\r\n* Fix dependency on filesystem on Linux (#137209)\r\n\r\nSimilar to: https://github.com/pytorch/pytorch/pull/134494\r\nWe are seeing come back of https://github.com/pytorch/pytorch/issues/133437 due to use of filesystem on Linux\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/137209\r\nApproved by: https://github.com/kit1980, https://github.com/malfet\r\n\r\n---------\r\n\r\nCo-authored-by: atalman \u003catalman@fb.com\u003e" }, { "commit": "783a6a424c30869f4b8ea8c686b4ce4991415e5c", "tree": "66f537fbae04f5ceff0a4a34b845d0ce26613431", "parents": [ "5375201dff598d2552b1c2f5ead027bd089415c6" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 15:19:35 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 15:19:35 2024 -0700" }, "message": "[MPS] Add regression test for `fft.fftfreq` (#137215)\n\n[MPS] Add regression test for `fft.fftfreq` (#135440)\r\n\r\nThe issue reported in #135223 was already solved in #128393. This PR adds a regression test for it.\r\n\r\nFixes #135223\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135440\r\nApproved by: https://github.com/ezyang\r\n\r\n(cherry picked from commit 09287e3af484421099d46ee56cc6ad7809402064)\r\n\r\nCo-authored-by: Roy Hvaara \u003croy@lightyear.no\u003e" }, { "commit": "5375201dff598d2552b1c2f5ead027bd089415c6", "tree": "13a8d4ecd8ac5431ed4a7bb3dfa7e78d32f4617f", "parents": [ "1de132ec9eab74d2a296002e4ce619e09abe2f43" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 14:54:41 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 14:54:41 2024 -0700" }, "message": "[MPS] Add missing dispatch to rshift.Tensor (#137212)\n\n[MPS] Add missing dispatch to rshift.Tensor (#135607)\r\n\r\nMissed it while working on https://github.com/pytorch/pytorch/pull/131813\r\nTest plan: `python -c \"import torch;print(torch.randint(100, 500, (64,), device\u003d\u0027mps\u0027) \u003e\u003e torch.tensor([3,], device\u003d\u0027mps\u0027))\"`\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135607\r\nApproved by: https://github.com/manuelcandales\r\n\r\n(cherry picked from commit 3bf6be457d40034aa4b603b7ea1b8977051221ed)\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "1de132ec9eab74d2a296002e4ce619e09abe2f43", "tree": "e545d1e0f88a311b324a7b9085f8eb8e4e2deb0b", "parents": [ "0b1b609ed7387a3e105c813b0d7e3f3c3df2bb36" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 14:52:55 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 14:52:55 2024 -0700" }, "message": "[MPS] Fix 5D+ reductions over negative dimentions (#137211)\n\n[MPS] Fix 5D+ reductions over negative dimentions (#136198)\r\n\r\nThis fixes bug introduced by https://github.com/pytorch/pytorch/pull/99856 that attempts to speed-up reduction for 5D+ tensor if trailing dimensions are all ones, but introduces crashes/off-by-one errors for wrapped dimensions\r\n\r\nAdded regresion test case to `TestMPS.test_sum`\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/136132\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136198\r\nApproved by: https://github.com/albanD\r\n\r\n(cherry picked from commit f6f1504d39c92f1ab2a1ee10a7da97745593151f)\r\n\r\nCo-authored-by: Nikita Shulga \u003cnikita.shulga@gmail.com\u003e" }, { "commit": "0b1b609ed7387a3e105c813b0d7e3f3c3df2bb36", "tree": "bb0e1889dfad22629b02001dc78dd6d586638161", "parents": [ "0b45af9c10612bc94ce53b9fcc22504629148e7e" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 14:51:50 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 14:51:50 2024 -0700" }, "message": "[NCCL] Don\u0027t override `waitUntilInitialized`\u0027s setting of `comm-\u003einitialized_` (#137210)\n\n[NCCL] Don\u0027t override `waitUntilInitialized`\u0027s setting of `comm-\u003einitialized_` (#136155)\r\n\r\n#133630 sets `initialized_` to `true` which causes previous wait codepaths to skip necessary waits, see also #https://github.com/pytorch/pytorch/issues/136151\r\n\r\nCC @shuqiangzhang @wconstab\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136155\r\nApproved by: https://github.com/fduwjj, https://github.com/kwen2501, https://github.com/c-p-i-o, https://github.com/shuqiangzhang\r\n\r\n(cherry picked from commit e3aa5e2f6410727efead4970a4ec30569cf76881)\r\n\r\nCo-authored-by: eqy \u003ceddiey@nvidia.com\u003e" }, { "commit": "0b45af9c10612bc94ce53b9fcc22504629148e7e", "tree": "6127539d6724686d943fc36a1da5c7e340f10b05", "parents": [ "1a0b166ba22be249ace1de014932803c60a17774" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 14:50:52 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 14:50:52 2024 -0700" }, "message": "Fix addmm silent correctness on aarch64 (#137208)\n\nFix addmm silent correctness on aarch64 (#136371)\r\n\r\nDo not dispatch to fast gemmv functions when alpha is not equal to 1\r\n\r\nAdd regression test to address the problem\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/136299\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136371\r\nApproved by: https://github.com/swolchok\r\n\r\n(cherry picked from commit c3e678382ba8eeeb3749b54e7b5256d7fb02cbd0)\r\n\r\nCo-authored-by: Nikita Shulga \u003cnikita.shulga@gmail.com\u003e" }, { "commit": "1a0b166ba22be249ace1de014932803c60a17774", "tree": "a8d3578e090b7c9c373461e9cc1a5a55867b19d0", "parents": [ "3a541ef8c27ba70eaef1f8612fd99fa4983cfcc7" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 13:12:27 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 13:12:27 2024 -0700" }, "message": "[ONNX] Add assertion nodes to ignoring list (#137214)\n\n[ONNX] Add assertion nodes to ignoring list (#135591)\r\n\r\nFixes #135419\r\n\r\nPS: there are 104 empty output nodes, I suggest we add them one by one when we run into them.\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135591\r\nApproved by: https://github.com/justinchuby\r\n\r\n(cherry picked from commit 492f064f15a56afd3839a3075f844632f9cd7964)\r\n\r\nCo-authored-by: titaiwangms \u003ctitaiwang@microsoft.com\u003e" }, { "commit": "3a541ef8c27ba70eaef1f8612fd99fa4983cfcc7", "tree": "89b056dcf5337007bf536cb1a8e9cb3d51e4efb7", "parents": [ "f8c4c252ca00cd07a3a37dcf60f7398f3d250de2" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Oct 02 12:57:34 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 12:57:34 2024 -0700" }, "message": "Clarify that `libtorch` API is C++17 compatible (#137206)\n\nClarify that `libtorch` API is C++17 compatible (#136471)\r\n\r\nAs it relies on some common C++17 primitives, such as `std::optional`\r\nReplace all docs references from C++14 to C++17\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/133205\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136471\r\nApproved by: https://github.com/kit1980, https://github.com/atalman\r\n\r\n(cherry picked from commit 4fd16dd8aa259cd75c9a6d2ddcd8171cd1ee8e28)\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "f8c4c252ca00cd07a3a37dcf60f7398f3d250de2", "tree": "7cc6aef155edf0d9a2d461377f302e03a059d6c8", "parents": [ "8af31b2e49d7544c378d2974807cfcfd234f32da" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Wed Oct 02 11:40:01 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Oct 02 11:40:01 2024 -0400" }, "message": "[Release only] Set WITH_PUSH when WITH_PUSH_ROCM is set (#137177)\n\n" }, { "commit": "8af31b2e49d7544c378d2974807cfcfd234f32da", "tree": "a1b1c68168ebc52897ea09d23c5a4e1703a7a22f", "parents": [ "8a71edcca517153dbb642890e8b21dd0b5d48152" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Tue Oct 01 18:40:25 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 01 18:40:25 2024 -0700" }, "message": "[RELEASE-ONLY Change] Push ROCm images on RC (#137148)\n\n* [RELEASE-ONLY Change] Push ROCm images on RC\r\n\r\nWe usually don\u0027t need to push docker images on RC, but for 2.5.0 release a cherry pick for ROCm actually modified docker images.\r\n\r\nDo this for ROCm only to be safe.\r\nAfter the release, think about what\u0027s the desired behavior and implement this in a more generic way.\r\n\r\n* Hardcode 2.5 in the tag" }, { "commit": "8a71edcca517153dbb642890e8b21dd0b5d48152", "tree": "759035d3c427a65aad59888b58dd093d596bbaa1", "parents": [ "058d3de7b9f5df5e6843e4013bd090f6d22a92ca" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Tue Oct 01 17:39:41 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Oct 01 17:39:41 2024 -0700" }, "message": "[RELEASE-ONLY CHANGES] Disable slow workflows (#136805)\n\n* Delete slow workflows\r\n\r\n* Create slow.yml" }, { "commit": "058d3de7b9f5df5e6843e4013bd090f6d22a92ca", "tree": "e266b190c0fa2c8ec6a870e64990f9ac20626c8d", "parents": [ "17d25897b20e6256a6b94d541e9f4f14a6e0cfcc" ], "author": { "name": "Animesh Jain", "email": "anijain@umich.edu", "time": "Mon Sep 30 18:02:31 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 18:02:31 2024 -0700" }, "message": "[dynamo] Do not treat user defined nn module attributes static for dynamic shape infra (#137025)\n\n[dynamo] Do not treat user defined nn module attributes static for dynamic shape infra (#136516)\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/136254\r\n\r\nTh regression was introduced in https://github.com/pytorch/pytorch/pull/132736 where originally we were trying to fix another regression. This PR and the offending PR together say - \"treat user defined nn module attributes as automatic dynamic, but for cudagraphs they will be considered static\". This avoid recompilations. This can lead to a cudagraph recording, which is ok. This also maintains the state before inline_inbuilt_nn_modules flag was introduced.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136516\r\nApproved by: https://github.com/williamwen42" }, { "commit": "17d25897b20e6256a6b94d541e9f4f14a6e0cfcc", "tree": "7a19fb7d2cdc4c701b7d651c71e1e7db4dbf5978", "parents": [ "70298e91f90aba76899c44f323f1dfbd3faa3a6f" ], "author": { "name": "Driss Guessous", "email": "32754868+drisspg@users.noreply.github.com", "time": "Mon Sep 30 18:01:24 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 18:01:24 2024 -0700" }, "message": "[FlexAttention] Fix output layout (#135882) (#136905)\n\nWe previously only supported the same v_head dim and + qk_head dim. When allowed for different head-dims I accidently kept the same query strides for the output. This PR fixes this bug as well it ensures that we always produce output in the same stride order as the input query.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135882\r\nApproved by: https://github.com/yanboliang, https://github.com/Chillee\r\n\r\n(cherry picked from commit ae02d663cdf493362699d2672ed7dc9019a7033b)" }, { "commit": "70298e91f90aba76899c44f323f1dfbd3faa3a6f", "tree": "41b280ca7af4b4052547ec6ca1d4c740ea13e6e0", "parents": [ "69ed7c709302c9de99884c47bc140cd6e17f0e05" ], "author": { "name": "Iris Z", "email": "31293777+wz337@users.noreply.github.com", "time": "Mon Sep 30 16:49:45 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 16:49:45 2024 -0700" }, "message": "[Cherry-pick][DSD] Fix distributed state dict full_state_dict option hang during set_state_dict (#135725) and Fix loading uneven full tensor into sharded state dict (#136365) (#136903)\n\n* [DSD] Fix distributed state dict full_state_dict option hang during set_state_dict (#135725)\r\n\r\nFix https://github.com/pytorch/pytorch/issues/134095\r\nThis fix distributed state dict full_state_dict option hang during set_state_dict. We switch `_distribute_tensors` in _state_dict_utils.py to use `DTensor.from_local` instead of `distribute_tensor` to support FSDP2+TP 2D strided sharding use case, as `distribute_tensor` cannot handle strided sharding yet. `distribute_tensor` incurs a scatter behind the scenes, while `DTensor.from_local` takes the local slice from the full tensor on each rank to create the DTensor (no collective). This means it\u0027s the user\u0027s responsibility to make sure the full_tensor from the full_state_dict is the same across all ranks.\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135725\r\nApproved by: https://github.com/fegin\r\n\r\n(cherry picked from commit 0cdc6a8dcd7e294b01d8914385bbe45e79c1770d)\r\n\r\n* [DSD] Fix loading uneven full tensor into sharded state dict (#136365)\r\n\r\nFix #136228.\r\n\r\nThis is a follow up on https://github.com/pytorch/pytorch/pull/135725. We need to pass shape and stride from the original dtensor, since for uneven case, `from_local` would calculate shape and stride assuming the tensor is evenly-sharded based on the local tensor.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136365\r\nApproved by: https://github.com/fegin\r\n\r\n(cherry picked from commit 637d5c4b7eb0fb82e019eed29efc0aa6ba92dc24)" }, { "commit": "69ed7c709302c9de99884c47bc140cd6e17f0e05", "tree": "68f5dcf3b34e9ca79a80612e6763996fdb4e3210", "parents": [ "d80f521ee2473763d6c2307931cea29b9c099782" ], "author": { "name": "Yifu Wang", "email": "yifu@fb.com", "time": "Mon Sep 30 15:54:05 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 15:54:05 2024 -0700" }, "message": "[SymmetricMemory] improve multicast initialization/fallback logic (#136894)\n\n[SymmetricMemory] improve multicast initialization/fallback logic (#136577)\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/136494\r\n\r\nCurrently, CUDASymmetricMemory::rendezvous() initializes a multicast address if multicast support is present. However, if we believe multicast support is present but cuMulticastCreate still fails for some reason, we do not fallback gracefully.\r\n\r\n- In addition to CUDART and driver version check, query CU_DEVICE_ATTRIBUTE_MULTICAST_SUPPORTED to determine multicast support for a rank/device.\r\n- Before initializing multicast for a block, ensure all ranks/devices have multicast support.\r\n- This is unlikely, but if cuMulticastCreate still fails on rank 0, print the corresponding driver error message as a warning, and gracefully skip multicast initialization for the block.\r\n- Introduced an environment variable (TORCH_SYMM_MEM_DISABLE_MULTICAST) to allow users to explicitly disable multicast support as a workaround.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136577\r\nApproved by: https://github.com/Chillee, https://github.com/eqy\r\n\r\n(cherry picked from commit d55eef5c596b3955dd8ee43c721b1c311dbab5e0)" }, { "commit": "d80f521ee2473763d6c2307931cea29b9c099782", "tree": "c62dd60fc14dbb50b0955b6c2b17018d2f9c9b72", "parents": [ "57717c8768eb74f7d214a860db4917ca4e6c8a3b" ], "author": { "name": "Jing Xu", "email": "jing.xu@intel.com", "time": "Tue Oct 01 07:51:36 2024 +0900" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 15:51:36 2024 -0700" }, "message": "fix requirements.txt installation failure issue on Windows (#136893)\n\n" }, { "commit": "57717c8768eb74f7d214a860db4917ca4e6c8a3b", "tree": "b15d4869ba7e93212eaae94676f6832fbc20bf4e", "parents": [ "550ed97a8960abbe3c8e375be5db9490acf7b1a4" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Mon Sep 30 14:56:13 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 14:56:13 2024 -0700" }, "message": "Fix lint (#137052)\n\nFix lint (#137023)\r\n\r\nBy migrating some of the workflows to Python-3.9 as 3.8 has been deprecated by https://github.com/pytorch/pytorch/pull/132138\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/137023\r\nApproved by: https://github.com/ZainRizvi, https://github.com/janeyx99, https://github.com/seemethere, https://github.com/kit1980, https://github.com/Skylion007\r\n\r\n(cherry picked from commit 40f80a70fab672e96a088796fa1b742a70b93070)\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "550ed97a8960abbe3c8e375be5db9490acf7b1a4", "tree": "bd760531b52ef1bccc6adbb5d84d762469b2e158", "parents": [ "051df20ac2360862a2830108eb10ba8e374a4455" ], "author": { "name": "eqy", "email": "eddiey@nvidia.com", "time": "Mon Sep 30 13:17:37 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 13:17:37 2024 -0700" }, "message": "[cuDNN][SDPA] cherrypick Support `attn_bias` in cuDNN (#130482) (#136885)\n\n[cuDNN][SDPA] Support `attn_bias` in cuDNN (#130482)\r\n\r\nCC @drisspg\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/130482\r\nApproved by: https://github.com/drisspg, https://github.com/Skylion007, https://github.com/malfet\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "051df20ac2360862a2830108eb10ba8e374a4455", "tree": "cbebeac0a8680ca68aa731efff68abbcc2bc1e6a", "parents": [ "bc421d456e83e54a0245e54ee3b4d2fbd1498ed2" ], "author": { "name": "Xinya Zhang", "email": "Xinya.Zhang@amd.com", "time": "Mon Sep 30 15:14:55 2024 -0500" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 30 13:14:55 2024 -0700" }, "message": "[ROCm] Update to AOTriton 0.7b (Cherry-picked) (#135869)\n\n* Always create seed and offset tensors on GPU memory.\r\n\r\n* Adjust fudge_factors for test_flash_attention_vs_math_ref_grads\r\n\r\n* Skip enable_gqa\u003dTrue tests\r\n\r\n* Fix cudagraph support for FA backend\r\n\r\n* Update the AOTriton FA API to meet hipGraph demands.\r\n\r\n* Enable test_fused_attention_vs_math_ref_grads_cudagraph and skip seq_len_q !\u003d seq_len_k when is_causal\u003dTrue\r\n\r\n* The main FA and ME tests passed after heavily hacking the fudge factors...\r\n\r\n* [SDPA] Add experimental support to Navi31\r\n\r\n* Changes aotriton_version.txt to 0.7b release\r\n\r\n* Make the fudge factors more explicit.\r\n\r\n* Code clean up.\r\n\r\n* Claim GQA is not supported on ROCM in can_use_flash_attention\r\n\r\n* Switch to .gz package\r\n\r\n* Skip failures on test/test_native_mha.py\r\n\r\n* Skip more GQA tests\r\n\r\n* Skip nn_functional_scaled_dot_product_attention related tests\r\n\r\n* Disable Efficient attention on fp32 + is_casual\u003dTrue\r\n\r\n* Revert \"Disable Efficient attention on fp32 + is_casual\u003dTrue\"\r\n\r\nThis reverts commit 36324a49d2c322146adbd678902fa32d008b8b8b.\r\n\r\nIt\u0027s not very effective and forcing MATH backend does not help. Need\r\nfurther investigations.\r\n\r\n* Add missing imports\r\n\r\n* Disable test_transformerencoderlayer and test_transformerdecoder\r\n\r\n* Fix two more problems\r\n\r\n* Fix lint\r\n\r\n* Re-enable test_transformers\r\n\r\n* Skip some tests in test_multiheadattention_fastpath_attn_mask on ROCM\r\n\r\n* fix lint\r\n\r\n* skip test_pointwise_associative_scan_tuple_reverse_True_combine_mode_pointwise_cuda on ROCm\r\n\r\n* skip more test_pointwise_associative_scan\r\n\r\n* Fix per suggestions from Nikita\r\n\r\n* Update skip reason of test_transformerencoderlayer\r\n\r\n* Add missing using" }, { "commit": "bc421d456e83e54a0245e54ee3b4d2fbd1498ed2", "tree": "1432b0c1ea67defcafb63e1c752f65036b989546", "parents": [ "aa574ab7e3fc49e6f2e6fef99e0322b6c39c665a" ], "author": { "name": "Gregory Comer", "email": "gregoryjcomer@gmail.com", "time": "Fri Sep 27 12:58:59 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 27 12:58:59 2024 -0700" }, "message": "[RELEASE ONLY CHANGES] Revert XNNPACK Update (#136522)\n\n* Revert \"Update generate-xnnpack-wrappers.py parsing to handle build identifier (#134724)\"\r\n\r\nThis reverts commit 679b8fe426babd40f8f9b013392df7a52901c926.\r\n\r\n* Revert \"Update PyTorch for XNNPACK 87ee0b4 (#134518)\"\r\n\r\nThis reverts commit 3b40b07efbc13422d59fff83ffaddc5141a5fa5e." }, { "commit": "aa574ab7e3fc49e6f2e6fef99e0322b6c39c665a", "tree": "1b74cb8c3885fa8d2d7c1888312655c494ad605b", "parents": [ "24bd87d5dd2dfb427150af0c3068b66496566b03" ], "author": { "name": "Joel Schlosser", "email": "75754324+jbschlosser@users.noreply.github.com", "time": "Fri Sep 27 12:44:04 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 27 09:44:04 2024 -0700" }, "message": "SDPA regression fix to work around high-precision by default (#136536)\n\nAdd option to configure reduced precision math backend for SDPA (#135964)\r\n\r\nSummary: Address https://github.com/pytorch/pytorch/issues/135778 by adding a global flag to configure whether using high precision or low precision for math backend of SDPA.\r\n\r\nTest Plan: buck2 run mode/opt //scripts/feikou/llm:run_attn_kernels\r\n\r\nDifferential Revision: D62625515\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135964\r\nApproved by: https://github.com/jbschlosser\r\n\r\n(cherry picked from commit 0a35986cdb72383c4e173abbdfc140ae2f98d5b1)\r\n\r\nCo-authored-by: Jianyu Huang \u003cjianyuhuang@meta.com\u003e" }, { "commit": "24bd87d5dd2dfb427150af0c3068b66496566b03", "tree": "253bff85c4bd3bc7d7ee59b516e371dd0897f94a", "parents": [ "6101aafa340fb02c47afcebf6d96b5d784bda8d6" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Thu Sep 26 18:59:00 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 26 18:59:00 2024 -0700" }, "message": "[Docs] fix inconsistent docs in conv1d, conv2d, and conv3d (#136813)\n\n[Docs] fix inconsistent docs in conv1d, conv2d, and conv3d (#135894)\r\n\r\nAddresses https://github.com/pytorch/pytorch/issues/135880\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135894\r\nApproved by: https://github.com/mikaylagawarecki, https://github.com/malfet\r\n\r\n(cherry picked from commit c9de2efde6ba0e1f15fe3ea99646855fdd9debaa)\r\n\r\nCo-authored-by: Sahan Paliskara \u003csahanp@meta.com\u003e" }, { "commit": "6101aafa340fb02c47afcebf6d96b5d784bda8d6", "tree": "baafd09a5f22db500b6b485441558f5b0254d719", "parents": [ "396413f05c9b27cc07e4deba855ca51545bb64f9" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Thu Sep 26 17:50:08 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 26 17:50:08 2024 -0700" }, "message": "[Update] Update note for Getting Started with PyTorch on Intel GPUs (#136731)\n\n[Update] Update note for Getting Started with PyTorch on Intel GPUs (#129946)\r\n\r\nremove the hardware and software prerequisites and set up env part.\r\nkeep the prerequisites section and link to pytorch prerequistes for intel gpus for driver install, intel support package install and env set up\r\nhttps://www.intel.com/content/www/us/en/developer/articles/tool/pytorch-prerequisites-for-intel-gpus.html\r\nUpdate the support for Intel Client GPU MTL-H\r\nUpdate inference \u0026 training examples\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/129946\r\nApproved by: https://github.com/seemethere\r\n\r\n(cherry picked from commit f3dd1721f48e503145669843a8fe90e0ad57d9f7)\r\n\r\nCo-authored-by: Zheng, Zhaoqiong \u003czhaoqiong.zheng@intel.com\u003e" }, { "commit": "396413f05c9b27cc07e4deba855ca51545bb64f9", "tree": "70ac58069f2577128eb384585b433c5a40847d40", "parents": [ "c25781c5d2422082df172b329c326e5b8999ed2a" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Thu Sep 26 16:09:37 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 26 16:09:37 2024 -0700" }, "message": "Fix ROCm skip decorator for test_ddp_tp and multiprocess UTs (#136161) (#136801)\n\nskip_if_rocm is used only in multiprocess case (when UT test class is a child of MultiProcessTestCase). Each individual process can exit with a skip code. If used for single process UT, it will cause the UT to fail as the process returns a non-zero exit code. Use skipIfRocm in single process UTs.\r\n\r\nTo avoid the above confusion, this PR renamed skip_if_rocm to skip_if_rocm_multiprocess.\r\n\r\nFixes #ISSUE_NUMBER\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136161\r\nApproved by: https://github.com/jithunnair-amd, https://github.com/kwen2501, https://github.com/fegin\r\n\r\nCo-authored-by: Prachi Gupta \u003cprachi.gupta@amd.com\u003e" }, { "commit": "c25781c5d2422082df172b329c326e5b8999ed2a", "tree": "21e0a8dd08a3bb51ae289f491eba78a5e87f141a", "parents": [ "ecd330669e64f72ca7ea01a1a6769f81e7d2767e" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Thu Sep 26 11:45:15 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 26 11:45:15 2024 -0700" }, "message": "Update current maintainers (#136769)\n\nUpdate current maintainers (#136672)\r\n\r\nThis file didn\u0027t had an overall in a few years so long overdue. Most of the credit goes to @orionr for gathering all of this info.\r\n\r\nThe main rules we followed:\r\n- No code contributor is removed, they\u0027re all placed as emeritus\r\n- Breakdown too big categories to make this document useful to know who to ping\r\n- No category where the code is still in the codebase is removed\r\n- We did not rework the categories (for example to be closer to module: labels) and leave that for later\r\n- All non-emeritus names are ordered by their number of comments on issues related to their topic\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136672\r\nApproved by: https://github.com/eqy, https://github.com/ezyang, https://github.com/seemethere, https://github.com/malfet\r\n\r\n(cherry picked from commit 2421344d8f582084b69b7b00fe0304b1c9732f65)\r\n\r\nCo-authored-by: albanD \u003cdesmaison.alban@gmail.com\u003e" }, { "commit": "ecd330669e64f72ca7ea01a1a6769f81e7d2767e", "tree": "b18b78139262a129797dfb3665c3414808fe8668", "parents": [ "1715708183d0f6c14511600fec1113fc6ffff0bf" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Thu Sep 26 00:05:27 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 26 00:05:27 2024 -0700" }, "message": "Constraint setuptools to 72.1.0 or older in requirements.txt (#136729)\n\nConstraint setuptools to 72.1.0 or older in requirements.txt (#136489)\r\n\r\nFIXES: https://github.com/pytorch/pytorch/issues/136541\r\n\r\nSetuptools\u003e\u003d74.0.0 has deprecated support for some functions in distutils, and so the builds run into error such as ```AttributeError: module \u0027distutils\u0027 has no attribute \u0027_msvccompiler\u0027```. Also, the pytorch builds have setuptools pin to 72.1.0 according to these PRs: https://github.com/pytorch/builder/pull/1995 and https://github.com/pytorch/builder/commit/89d9a8cf6f4357d5d523b8d7c46351ac913e1309. So, until there is a fix to change the function usage in accordance with latest setuptools, the 72.1.0 version works fine.\r\n\r\nAlso observed in CI jobs: https://github.com/pytorch/pytorch/actions/runs/10979326524\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136489\r\nApproved by: https://github.com/malfet\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e\r\n(cherry picked from commit 8d65d9f11bbbcb0ea4152510ffe9784135cfacc8)\r\n\r\nCo-authored-by: ratnampa \u003cratnam.parikh@intel.com\u003e" }, { "commit": "1715708183d0f6c14511600fec1113fc6ffff0bf", "tree": "bc12ee45a75090f41742cdc55628e9101cec6157", "parents": [ "2e2c00f74cc9443f8c10f71f029b356c79ea06ee" ], "author": { "name": "Yanbo Liang", "email": "ybliang8@gmail.com", "time": "Wed Sep 25 23:10:28 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 23:10:28 2024 -0700" }, "message": "Revert \"Trace fwd graph under no_grad mode #134872\" (#136734)\n\nThis reverts commit aca4d6b891ac1b7296195283edb8b64c828a6351." }, { "commit": "2e2c00f74cc9443f8c10f71f029b356c79ea06ee", "tree": "a88c4a7beea8cb10b461d51dc3799d6d6abb5e47", "parents": [ "cbe476a5a7f0b4dcc1e5e757cabef5833dc4f46c" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Sep 25 18:33:27 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 18:33:27 2024 -0700" }, "message": "Make test_skip_data_serialization regex more flexible (#136710)\n\nMake test_skip_data_serialization regex more flexible (#136580)\r\n\r\nSome CI machines seem to throw \"Can\u0027t get local object\" rather than\r\n\"Can\u0027t pickle local object\".\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136580\r\nApproved by: https://github.com/mikaylagawarecki\r\n\r\n(cherry picked from commit a0c76ea8533a4327fa36dfeeaf0863749d5e6dca)\r\n\r\nCo-authored-by: Jez Ng \u003cjezng@fb.com\u003e" }, { "commit": "cbe476a5a7f0b4dcc1e5e757cabef5833dc4f46c", "tree": "9eb2d055ca5f76a81f063577297bc427d699fb93", "parents": [ "4b030d47b1bb1f4fb34fa949be4be2c08808f0e9" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Sep 25 18:18:23 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 18:18:23 2024 -0700" }, "message": "Disable iOS workflow (#136706)\n\nDisable iOS workflow (#136571)\r\n\r\nSee https://github.com/pytorch/pytorch/issues/136284\r\nIt\u0027s been broken for more than a week and it does not seem like anyone cares about fixing it.\r\nOnce it\u0027s landed I\u0027ll reassigned the issue on `oncall: mobile`\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136571\r\nApproved by: https://github.com/huydhn, https://github.com/kit1980\r\n\r\n(cherry picked from commit 5340feb8aa97fe0a70251a52f2c9b570adf7dbe0)\r\n\r\nCo-authored-by: Nikita Shulga \u003c2453524+malfet@users.noreply.github.com\u003e" }, { "commit": "4b030d47b1bb1f4fb34fa949be4be2c08808f0e9", "tree": "5436c2ba3a2dc1578f168611d6ae3a98b5ecd672", "parents": [ "9b80ddecd6c580c7b40576ec900c37fc1cbdf803" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Wed Sep 25 18:03:12 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 18:03:12 2024 -0700" }, "message": "[RELEASE-ONLY CHANGES] Don\u0027t push to https://ghcr.io/ (#136703)\n\nDon\u0027t do push to https://ghcr.io/ on release branch: we don\u0027t need it and it fails with \"unauthorized: unauthenticated: User cannot be authenticated with the token provided\"." }, { "commit": "9b80ddecd6c580c7b40576ec900c37fc1cbdf803", "tree": "8479c64852f2841411e3ed28cd5b258f12247489", "parents": [ "6e86793f7591209b6fac9b4a0ae9878b84fe42a6" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Sep 25 17:45:39 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 17:45:39 2024 -0700" }, "message": "Fix hardcoded ROCm paths in `Caffe2Targets.cmake` (#136700)\n\nFix hardcoded ROCm paths in `Caffe2Targets.cmake` (#136283)\r\n\r\nFixes #131701\r\n\r\nUse CMake imported targets more consistently to eliminate hardcode paths.\r\n\r\nHere is the new relevant sections of Caffe2Targets.cmake:\r\n```\r\nset_target_properties(c10_hip PROPERTIES\r\n INTERFACE_INCLUDE_DIRECTORIES \"${_IMPORT_PREFIX}/include\"\r\n INTERFACE_LINK_LIBRARIES \"c10;hip::amdhip64\"\r\n)\r\n```\r\n\r\n```\r\nset_target_properties(torch_hip PROPERTIES\r\n INTERFACE_COMPILE_DEFINITIONS \"USE_C10D_NCCL\"\r\n INTERFACE_COMPILE_OPTIONS \"-fPIC;-D__HIP_PLATFORM_AMD__\u003d1;-DCUDA_HAS_FP16\u003d1;-DUSE_ROCM;-D__HIP_NO_HALF_OPERATORS__\u003d1;-D__HIP_NO_HALF_CONVERSIONS__\u003d1;-DTORCH_HIP_VERSION\u003d602;-Wno-shift-count-negative;-Wno-shift-count-overflow;-Wno-duplicate-decl-specifier;-DCAFFE2_USE_MIOPEN;-DTHRUST_DEVICE_SYSTEM\u003dTHRUST_DEVICE_SYSTEM_HIP;-std\u003dc++17;-DHIPBLAS_V2;-DHIP_NEW_TYPE_ENUMS\"\r\n INTERFACE_INCLUDE_DIRECTORIES \"${_IMPORT_PREFIX}/include\"\r\n INTERFACE_LINK_LIBRARIES \"c10_hip;torch_cpu_library;hip::amdhip64;MIOpen;hiprtc::hiprtc;roc::hipblaslt;roc::hipblas;hip::hipfft;hip::hiprand;roc::hipsparse;roc::hipsolver\"\r\n)\r\n```\r\n\r\nHIPCUB dependency was not actually used; which is why it is removed here as the imported target had undesirable side effects.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136283\r\nApproved by: https://github.com/jeffdaily, https://github.com/Skylion007, https://github.com/jithunnair-amd, https://github.com/atalman\r\n\r\n(cherry picked from commit e8f1dd6ba0675d3e11808e5198b0d927278a6f91)\r\n\r\nCo-authored-by: Nichols A. Romero \u003cnick.romero@amd.com\u003e" }, { "commit": "6e86793f7591209b6fac9b4a0ae9878b84fe42a6", "tree": "b644b01d79055b982e688aa48152aa55e832560a", "parents": [ "7c550fea95aa7523272a90f7f575708b1c44a508" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Wed Sep 25 17:17:30 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 17:17:30 2024 -0700" }, "message": "[ROCm] upgrade ROCm CI builds to py3.10 (#134108) (#136696)\n\nUpgrade ROCm CI builds to py3.10\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/134108\r\nApproved by: https://github.com/jeffdaily, https://github.com/jithunnair-amd, https://github.com/atalman\r\n\r\nCo-authored-by: Jack Taylor \u003cjack.taylor@amd.com\u003e" }, { "commit": "7c550fea95aa7523272a90f7f575708b1c44a508", "tree": "8088c4cc96fa33cc43b059bd7c8969018d21ff27", "parents": [ "7a00785c2354784ad6030c94076f506a9a322710" ], "author": { "name": "Jeff Daily", "email": "jeff.daily@amd.com", "time": "Wed Sep 25 13:42:27 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 13:42:27 2024 -0700" }, "message": "[ROCm][CI] upgrade CI to ROCm 6.2 (#132555) (#136467)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/132555\r\nApproved by: https://github.com/pruthvistony, https://github.com/malfet" }, { "commit": "7a00785c2354784ad6030c94076f506a9a322710", "tree": "185dc792fce040f50b00bdd658ff18db4fe32fc5", "parents": [ "4e6a99e5f39df87b509917bf07ef61426bdda9f5" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Wed Sep 25 13:12:47 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 16:12:47 2024 -0400" }, "message": "[ROCm] Cherry-pick unit test fixes to release/2.5 (#136557)\n\n* [ROCm] skip test_fp8_cast_and_t on non-MI300 machines (#135917)\r\n\r\nFixes #ISSUE_NUMBER\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135917\r\nApproved by: https://github.com/malfet\r\n\r\n(cherry picked from commit 6cdc70bccd369d0d7ec4b54d921c91d62bd931f7)\r\n\r\n* Skip pointwise associative scan tests due to regression (changes based on PR https://github.com/pytorch/pytorch/pull/135995)\r\n\r\n* Cherry-pick fix from https://github.com/pytorch/pytorch/pull/135702\r\n\r\n---------\r\n\r\nCo-authored-by: Prachi Gupta \u003cprachi.gupta@amd.com\u003e\r\nCo-authored-by: Jithun Nair \u003cjithun.nair@amd.com\u003e" }, { "commit": "4e6a99e5f39df87b509917bf07ef61426bdda9f5", "tree": "b56a8056cb7647279a1c937d649a520ee3bc3c6f", "parents": [ "dd73223b90508ae9c3322f0eeed006119198d732" ], "author": { "name": "Shan19900305", "email": "62477751+Shan19900305@users.noreply.github.com", "time": "Thu Sep 26 03:26:29 2024 +0800" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 12:26:29 2024 -0700" }, "message": "fix stride compare failed when size value equal to one in ForeachUtils.h (#136426)\n\nfix stride compare failed when size value equal to one in ForeachUtils.h (#134546)\r\n\r\nWhen size value equal to one, tensor strides value need be skipped to compare.\r\n@ezyang\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/134546\r\nApproved by: https://github.com/janeyx99" }, { "commit": "dd73223b90508ae9c3322f0eeed006119198d732", "tree": "e3f94c89cacbdad295eefbb51db9d545ba2a4b73", "parents": [ "c5e5254a796446275ed9ba8c7842c50bdee164e7" ], "author": { "name": "Jack Taylor", "email": "108682042+jataylo@users.noreply.github.com", "time": "Wed Sep 25 18:56:24 2024 +0000" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 25 11:56:24 2024 -0700" }, "message": "[ROCm] [BUGFIX] Re-enable rocm-specific tuning parameters v2 (#133852) (#136139)\n\nSmall bug fix - https://github.com/pytorch/pytorch/pull/124592 replaced the torch.version.hip with device_props but made a mistake in porting the original logic.\r\n\r\nThe original code was:\r\n`if torch.version.hip is not None:`\r\n\r\nWhich was incorrectly replaced by:\r\n`if self.device_props.type !\u003d \"hip\":`\r\n\r\nAnother occurence of https://github.com/pytorch/pytorch/pull/130617\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/133852\r\nApproved by: https://github.com/masnesral, https://github.com/malfet\r\n\r\n(cherry picked from commit da587de9cb7cdbcd69f65783cd7a9589198bd6f6)" }, { "commit": "c5e5254a796446275ed9ba8c7842c50bdee164e7", "tree": "9f1136b4a40aa3c1d4ebd5cbbc3ade2345802ba6", "parents": [ "ffed7b71e8ce6eb3e73e925511b2bd84df2f5037" ], "author": { "name": "Yu, Guangye", "email": "106960996+guangyey@users.noreply.github.com", "time": "Tue Sep 24 10:58:37 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Sep 24 13:58:37 2024 -0400" }, "message": "Fix xpu memory stats error (#135818) (#136420)\n\n# Motivation\r\nfix https://github.com/pytorch/pytorch/issues/135726\r\nAfter merging two free blocks, I made a stupid mistake of ignoring the correct size to decrease the active memory size, which should be the original block size instead of the merged block size.\r\n\r\n# Additional Context\r\nAdd a UT to guard this scenario.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135818\r\nApproved by: https://github.com/EikanWang\r\n\r\n(cherry picked from commit e6b68359d7c86aff25eefe77e0774c02b38f44b4)" }, { "commit": "ffed7b71e8ce6eb3e73e925511b2bd84df2f5037", "tree": "1cfd37dd1b8be28fe32f212eef681b9db4fbc7bd", "parents": [ "becdf8ae4fd221d90c8d02a82da50f34dc56e65e" ], "author": { "name": "xiangdong", "email": "40376367+zxd1997066@users.noreply.github.com", "time": "Fri Sep 20 15:47:42 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:47:42 2024 -0700" }, "message": "Fix dynamo benchmark skip logic for cpu device (#135193) (#135793)\n\nFixes #132380, adjust torchbench and huggingface skip models list, then we can remove `--no-skip` when running benchmarks on 3 suites.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135193\r\nApproved by: https://github.com/chuanqi129, https://github.com/jansel\r\n\r\n(cherry picked from commit 7ec17b49cf89cfeb97272a1baddcc30fa6fa66d8)" }, { "commit": "becdf8ae4fd221d90c8d02a82da50f34dc56e65e", "tree": "d732da0c6645bdbd629b5a18a63cb7983cc5aeb9", "parents": [ "fb276d2652b10c8b0c63d1b31d5e52bc94996ca9" ], "author": { "name": "jiayisunx", "email": "jiayi.sun@intel.com", "time": "Sat Sep 21 06:45:26 2024 +0800" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:45:26 2024 -0700" }, "message": "[Inductor] Increase multiplier to 3 for Inductor AMP FP16 benchmark correctness check (#135932) (#136262)\n\nFix https://github.com/pytorch/pytorch/issues/135657.\r\nAligned with AMP BF16, using multiplier 3 for Inductor AMP FP16 benchmark correctness check\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135932\r\nApproved by: https://github.com/CaoE, https://github.com/jgong5, https://github.com/jansel" }, { "commit": "fb276d2652b10c8b0c63d1b31d5e52bc94996ca9", "tree": "68f1a98daf2a4e5792199b466676a4c5625a81de", "parents": [ "b7de7932fd446d96be07267effc2fbac9450e13f" ], "author": { "name": "Justin Chu", "email": "justinchuby@users.noreply.github.com", "time": "Fri Sep 20 15:44:05 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:44:05 2024 -0700" }, "message": "[ONNX] Fix numpy method to return the correct type (#136162) (#136203)\n\n[ONNX] Fix numpy method to return the correct type (#136162)\r\n\r\nPrevious implementation of the `numpy()` method returns `fp64` when the tensor is `fp32`. This is unexpected but seems to be caused by calling `__array__(dtype\u003dNone)` on the numpy array. I updated the implementation to implement the `numpy()` method explicitly and added tests to guard the behavior.\r\n\r\nThis needs to be cherry-picked into torch 2.5\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/136162\r\nApproved by: https://github.com/gramalingam, https://github.com/xadupre\r\n\r\n(cherry picked from commit 67b14ce8bd9d4d0ad1920e57bc148644775646ac)" }, { "commit": "b7de7932fd446d96be07267effc2fbac9450e13f", "tree": "368abd3dd51d7b5ebb4d35d968e0d1c0b71d68c5", "parents": [ "1954439802e10eff7c6cc03800f23466cf2d6844" ], "author": { "name": "Feng Yuan", "email": "feng1.yuan@intel.com", "time": "Fri Sep 20 15:39:53 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:39:53 2024 -0700" }, "message": "Update torch-xpu-ops pin (ATen XPU implementation) (#135833)\n\nUpdate torch-xpu-ops pin (ATen XPU implementation) (#135647)\r\n\r\nRelease cycle for PyTorch 2.5\r\n1. Fixing runtime error on Windows: Fail to load torch_xpu_ops_unary_binary_kernels.dll as the bin size is large.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135647\r\nApproved by: https://github.com/EikanWang" }, { "commit": "1954439802e10eff7c6cc03800f23466cf2d6844", "tree": "0e9364c49adf2367ec87f9c6dc4b23dab6858c68", "parents": [ "3920988456897eac63003ad84f52157d61fbaef2" ], "author": { "name": "Cao E", "email": "e.cao@intel.com", "time": "Sat Sep 21 06:28:26 2024 +0800" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:28:26 2024 -0700" }, "message": "Update document for autocast on CPU (#136082)\n\nupdate document for autocast on CPU" }, { "commit": "3920988456897eac63003ad84f52157d61fbaef2", "tree": "b3a8ae6728f44aa4ebcf446dd8e19b4432342851", "parents": [ "1db2a6562c9a6ef64e33766c0466ed2c66430adf" ], "author": { "name": "Shubham Bhokare", "email": "32080845+shubhambhokare1@users.noreply.github.com", "time": "Fri Sep 20 15:25:25 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:25:25 2024 -0700" }, "message": "[Cherry-pick] [ONNX] Update fake mode usage in onnx docs (#135512) and Drop final None values as inputs for nodes in exporter graph (#135520) (#136005)\n\n* [ONNX] Update fake mode usage in onnx docs (#135512)\r\n\r\nUpdate fake mode usage in onnx docs\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135512\r\nApproved by: https://github.com/justinchuby\r\n\r\nCo-authored-by: Justin Chu \u003cjustinchuby@users.noreply.github.com\u003e\r\n(cherry picked from commit 66db61f0d1b55323af958c3401d74842dc00b0c7)\r\n\r\n* [ONNX] Drop final None values as inputs for nodes in exporter graph (#135520)\r\n\r\nWhen value for an optional input is not provided, it is defaulted to `None`, which gets translates to \"\" in the onnx graph. To avoid this, if we have a list of inputs and the final few are all `None`, strip them in the graph.\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135520\r\nApproved by: https://github.com/justinchuby\r\n\r\nCo-authored-by: Justin Chu \u003cjustinchuby@users.noreply.github.com\u003e\r\n(cherry picked from commit e2f9a83b85af1bca54d97d2cc1259d72bf9579e5)" }, { "commit": "1db2a6562c9a6ef64e33766c0466ed2c66430adf", "tree": "ddc69f34bf44b59451f6d16d54f83d4c6896bebc", "parents": [ "4b5bf41476c0b4ce0e5cb1392970677cfca8c343" ], "author": { "name": "Aaron Gokaslan", "email": "aaronGokaslan@gmail.com", "time": "Fri Sep 20 15:24:10 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:24:10 2024 -0700" }, "message": "[DCP] Fixes the stateless optimizer issue of distributed state_dict (… (#136000)\n\n[DCP] Fixes the stateless optimizer issue of distributed state_dict (#135535)\r\n\r\nSome optimizers don\u0027t have states that can cause get_state_dict/set_state_dict behave incorrectly. This PR fixes the issues.\r\n\r\nfixes: https://github.com/pytorch/pytorch/issues/133415\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135535\r\nApproved by: https://github.com/wz337\r\n\r\nCo-authored-by: Chien-Chin Huang \u003cchienchin@fb.com\u003e" }, { "commit": "4b5bf41476c0b4ce0e5cb1392970677cfca8c343", "tree": "b13cb9891d67712e14ad373cb38fb85bfae54c68", "parents": [ "a889c8549817ea34879968d7b3351f58b3f4e539" ], "author": { "name": "Chunyuan WU", "email": "chunyuan.wu@intel.com", "time": "Sat Sep 21 06:20:16 2024 +0800" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Fri Sep 20 15:20:16 2024 -0700" }, "message": "[inductor] [cpp] fix the input contiguous check in max-autotune (#135561)\n\n[inductor] [cpp] fix the input contiguous check in max-autotune (#134982)\r\n\r\n## Description\r\nFixes the FP32 accuracy failure of `resmlp_12_224` and BF16 accuracy failure of `volo_d1_224` in timm.\r\n\r\nIn this PR, we check whether input is contiguous using the following way:\r\nIf it has `FixedLayout`, we know the accurate strides. For `FlexibleLayout`, if its data is a `ComputedBuffer`, we could get the fill order of the buffer to decide whether it\u0027s contiguous. For the other cases, we won\u0027t use GEMM template as we can\u0027t infer whether it\u0027s contiguous.\r\n\r\n## Additional context\r\nThe current GEMM template only supports this case: `input.get_stride()[-1] \u003d\u003d 1`. In `resmlp_12_224`, when we run into this check, the layout of `input` is a `FlexibleLayout`. The reason is that when realizing the input which is a `View` IR, the `convert_to_reinterpret_view` call fails:\r\nhttps://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/ir.py#L4712-L4715\r\n\r\nAnd it finally runs into this `copy_input` and returns a `FlexibleLayout`.\r\nhttps://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/ir.py#L4722\r\n\r\nWhen checking its stride, this `FlexibleLayout` indeed satisfies `input.get_stride()[-1] \u003d\u003d 1` but it is later decided as a `FixedLayout` with `size \u003d (3072, 196), stride \u003d (1, 3072)`, which is not supported by the GEMM template, thus causing accuracy issue in this model.\r\nThe `FlexibleLayout` is converted to `FixedLayout` during [CppPackedGemmTemplate.add_choices](https://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/mkldnn_lowerings.py#L1051) which calls [slice_nd](https://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/codegen/cpp_template_kernel.py#L150) when rendering the kernel (`slice_nd(X)`). When creating the `SliceView` IR, [as_storage_and_layout](https://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/ir.py#L2288) invokes\r\n[decide_layout](https://github.com/pytorch/pytorch/blob/d14fe3ffeddff743af09ce7c8d91127940ddf7ed/torch/_inductor/ir.py#L2135) and converts it to a `FixedLayout` with `size \u003d (3072, 196), stride \u003d (1, 3072)`.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/134982\r\nApproved by: https://github.com/jgong5, https://github.com/leslie-fang-intel, https://github.com/jansel" }, { "commit": "a889c8549817ea34879968d7b3351f58b3f4e539", "tree": "c997524be397a48c068d435dee4042b6b85b4fce", "parents": [ "813e06461c231efb17e46934eafe8c37b5e958ab" ], "author": { "name": "jiayisunx", "email": "jiayi.sun@intel.com", "time": "Fri Sep 20 08:03:24 2024 +0800" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 19 17:03:24 2024 -0700" }, "message": "[Inductor] simplify indexing_exprs in LoopBody._init_with_copy (#135574) (#135935)\n\nThis PR uses `var_ranges` information to simplify `indexing_exprs` in `LoopBody._init_with_copy` to to reduce occurrences of `FloorDiv` and `ModularIndexing` in the `indexing_exprs`.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135574\r\nApproved by: https://github.com/jgong5, https://github.com/leslie-fang-intel, https://github.com/jansel" }, { "commit": "813e06461c231efb17e46934eafe8c37b5e958ab", "tree": "613abf15d403920ee2c284c7dc7605312a55bec8", "parents": [ "9887030485f2da7a139c5fb1be644d9cfd68cd16" ], "author": { "name": "Justin Chu", "email": "justinchuby@users.noreply.github.com", "time": "Thu Sep 19 16:58:05 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 19 16:58:05 2024 -0700" }, "message": "[ONNX] Fix symbolic values and numpy implementation (#135786) (#135868)\n\n1. Remove `__eq__` to make `SymbolicTensor` hashable and test for that\r\n2. Update the `__array__` method so that it works for tensor on GPU\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/135700\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135786\r\nApproved by: https://github.com/titaiwangms" }, { "commit": "9887030485f2da7a139c5fb1be644d9cfd68cd16", "tree": "2981867ae01d3d0ce1cdc58ff94ef4074db27f6e", "parents": [ "9e315fef22bc63b892ac90ae97b6eacd7227fb75" ], "author": { "name": "Sergii Dymchenko", "email": "sdym@meta.com", "time": "Thu Sep 19 16:50:49 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 19 16:50:49 2024 -0700" }, "message": "Revert \"[fx] Bypass custom __setattr__ in Node.__init__ (#135079)\" (#… (#135625)\n\nRevert \"[fx] Bypass custom __setattr__ in Node.__init__ (#135079)\" (#135562)\r\n\r\nThis reverts commit 66da3b3b2acacb116a9b23e91b24934830eaf6b8.\r\n\r\n#135079 breaks internal tests and needs to be reverted. Revert with mergebot doesn\u0027t work as this PR is technically part of the stack, but, according to @jansel, it should be possible to revert it individually.\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135562\r\nApproved by: https://github.com/jansel, https://github.com/seemethere\r\n\r\nCo-authored-by: Ivan Zaitsev \u003civanzaitsev@fb.com\u003e" }, { "commit": "9e315fef22bc63b892ac90ae97b6eacd7227fb75", "tree": "1cc41da5c155c0907a7bb69f6e2bb32da45be789", "parents": [ "6b14e6cfddc62f71a65f250c58d2482fdf2126b4" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Wed Sep 18 11:17:52 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 18 11:17:52 2024 -0700" }, "message": "Revert \"[Release only] Temporary disable triton xpu build\" (#136276)\n\nRevert \"[Release only] Temporary disable triton xpu build (#136206)\"\r\n\r\nThis reverts commit 6b14e6cfddc62f71a65f250c58d2482fdf2126b4." }, { "commit": "6b14e6cfddc62f71a65f250c58d2482fdf2126b4", "tree": "f0aa9caf55e2f01cd340c172fb94d30b070a073c", "parents": [ "828d686e1ccb09f1273f2c924680cf3d8b65082c" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Tue Sep 17 15:12:09 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Sep 17 12:12:09 2024 -0700" }, "message": "[Release only] Temporary disable triton xpu build (#136206)\n\n" }, { "commit": "828d686e1ccb09f1273f2c924680cf3d8b65082c", "tree": "1cc41da5c155c0907a7bb69f6e2bb32da45be789", "parents": [ "612fc7c447ce2e4b7fec1116433bdc3f1dda4d68" ], "author": { "name": "Ti-Tai Wang", "email": "titaiwang@microsoft.com", "time": "Thu Sep 12 09:50:58 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 12 09:50:58 2024 -0700" }, "message": "[ONNX] Fix scaled_dot_product_attention with float scale (#135710)\n\nFixes #125158\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135594\r\nApproved by: https://github.com/justinchuby\r\n\r\n(cherry picked from commit e48ee2cf50d86d87ef7c7d0839267dbed4903ebf)" }, { "commit": "612fc7c447ce2e4b7fec1116433bdc3f1dda4d68", "tree": "306cd67ea3ad75d97bddda621ae04db825ea250f", "parents": [ "cea562006e5f00d13f43c093b51256ebee91f262" ], "author": { "name": "Justin Chu", "email": "justinchuby@users.noreply.github.com", "time": "Thu Sep 12 09:49:45 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Thu Sep 12 09:49:45 2024 -0700" }, "message": "[ONNX] Improves documentation of ONNX exporter (#135526)\n\nThe PR updates the documentation to reflect the changes introduced in pytorch 2.5 and related to onnx exporter.\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135372\r\nApproved by: https://github.com/justinchuby\r\n\r\nCo-authored-by: Justin Chu \u003cjustinchuby@users.noreply.github.com\u003e\r\n(cherry picked from commit 5e145861f23370d342125118dd9b0dcba26147db)\r\n\r\nCo-authored-by: Xavier Dupré \u003cxadupre@users.noreply.github.com\u003e" }, { "commit": "cea562006e5f00d13f43c093b51256ebee91f262", "tree": "37f774ba3df67809767ca957c82d33dc23491bc3", "parents": [ "ba275025016eda19869405fdb78aeb1ec163fb55" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Wed Sep 11 15:03:44 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Wed Sep 11 15:03:44 2024 -0400" }, "message": "[Cherry-Pick] Bump triton pin and release version, revert temporary changes to build from pin (#135613)\n\n* Revert \"[RELEASE-ONLY CHANGES] Temp changes to build triton from pin for first RC (#135517)\"\r\n\r\nThis reverts commit 4a3dabd67f8ce63f2fc45f278421cca3cc532cfe.\r\n\r\n* Build triton from release branch\r\n\r\n* triton_pin\r\n\r\n* fix\r\n\r\n* Bump triton xpu pin and release version (#135638)\r\n\r\nSimilar with https://github.com/pytorch/pytorch/pull/135627\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135638\r\nApproved by: https://github.com/atalman\r\n\r\n---------\r\n\r\nCo-authored-by: chuanqiw \u003cchuanqi.wang@intel.com\u003e" }, { "commit": "ba275025016eda19869405fdb78aeb1ec163fb55", "tree": "1d9272a4326ff8c873b17bc9a613006343740a41", "parents": [ "4a3dabd67f8ce63f2fc45f278421cca3cc532cfe" ], "author": { "name": "pytorchbot", "email": "soumith+bot@pytorch.org", "time": "Tue Sep 10 11:43:16 2024 -0700" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Tue Sep 10 14:43:16 2024 -0400" }, "message": "Use upload-artifact@v4.4.0 for create_release.yml (#135534)\n\nUse upload-artifact@v4.4.0 for create_release.yml (#135528)\r\n\r\nFixes failure: https://github.com/pytorch/pytorch/actions/runs/10780281005/job/29895846007\r\n\r\nDue broken sync\r\n```\r\nactions/upload-artifact@v2\r\nand\r\nactions/download-artifact@v4.1.7\r\n```\r\n\r\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135528\r\nApproved by: https://github.com/kit1980, https://github.com/malfet\r\n\r\n(cherry picked from commit 9b764491e32755d64efd4e7ff1c90cf587f3e3ff)\r\n\r\nCo-authored-by: atalman \u003catalman@fb.com\u003e" }, { "commit": "4a3dabd67f8ce63f2fc45f278421cca3cc532cfe", "tree": "9a0871b7963c2109eb538cc3de3f41cc24b29672", "parents": [ "e13069627086d0c720a96f1e513510e1edc14071" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Mon Sep 09 15:08:05 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 09 15:08:05 2024 -0400" }, "message": "[RELEASE-ONLY CHANGES] Temp changes to build triton from pin for first RC (#135517)\n\n[Release only] Temp changes to build triton from pin" }, { "commit": "e13069627086d0c720a96f1e513510e1edc14071", "tree": "d9f13d113bacb37e461748187cd251f079f0ef4a", "parents": [ "b7eb7256fb9a48d1fc452608986b64688b6469fa" ], "author": { "name": "Andrey Talman", "email": "atalman@fb.com", "time": "Mon Sep 09 14:05:58 2024 -0400" }, "committer": { "name": "GitHub", "email": "noreply@github.com", "time": "Mon Sep 09 14:05:58 2024 -0400" }, "message": "[RELEASE-ONLY CHANGES] Branch Cut for Release 2.5 (#135506)\n\n* [RELEASE-ONLY CHANGES] Branch Cut for Release 2.5\r\n\r\n* fix_lint" }, { "commit": "b7eb7256fb9a48d1fc452608986b64688b6469fa", "tree": "44ad7a50a15b46ec4dbeb9aae6dc0a28950bd364", "parents": [ "c1ae78be9283822c8541e271fe3f290ae12d4837" ], "author": { "name": "Alexander Kurakin", "email": "kuraga333@mail.ru", "time": "Mon Sep 09 03:16:09 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Mon Sep 09 03:16:11 2024 +0000" }, "message": "docs: `torch.nn.utils.rnn.pack_padded_sequence`: docs improve (#135417)\n\ndocs: `torch.nn.utils.rnn.pack_padded_sequence`: docs improve\n\n/cc @mikaylagawarecki\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135417\nApproved by: https://github.com/ezyang\n" }, { "commit": "c1ae78be9283822c8541e271fe3f290ae12d4837", "tree": "ff22c110eed1e612404cc4bb1984887bcb121ffc", "parents": [ "defb515306fc53ec62e92937a5a76fa5cbc05b84" ], "author": { "name": "Xu Han", "email": "xu.han@outlook.com", "time": "Mon Sep 09 03:10:54 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Mon Sep 09 03:10:54 2024 +0000" }, "message": "[inductor] calibration inductor windows uts (18/N) (#135449)\n\nskip test_quantized_* UTs of `test/inductor/test_cpu_select_algorithm.py`.\nWindows inductor don\u0027t support quantize so far.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135449\nApproved by: https://github.com/ezyang\n" }, { "commit": "defb515306fc53ec62e92937a5a76fa5cbc05b84", "tree": "ea5bfb0864c9f8102c487acf261ef91f030ef0c5", "parents": [ "31c4e0d37d8efc37a0697159e5b9121ec34d5141" ], "author": { "name": "yuqingj", "email": "yuqingj@meta.com", "time": "Sat Sep 07 20:41:46 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 21:00:41 2024 +0000" }, "message": "[NJT]Add permute ops support (#135336)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135336\nApproved by: https://github.com/davidberard98\n" }, { "commit": "31c4e0d37d8efc37a0697159e5b9121ec34d5141", "tree": "e4e295d76be3bd448e40c67aeb4320fffadc53eb", "parents": [ "53290ca00b83cf84b1074e0d9e0b7096885a93c4" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Sat Sep 07 19:54:41 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 18:02:36 2024 +0000" }, "message": "[inductor] Cleanup analysis done at lowering time (#135412)\n\nBefore this we would take multiple passes over the body of each IRNode as we did lowering. This combines most analysis into `OpCounterCSE` so it can be done in a single pass.\n\nBefore:\n![image](https://github.com/user-attachments/assets/0047db09-4258-4491-a9a6-b078e183092a)\n\nAfter:\n![image](https://github.com/user-attachments/assets/1e03adcb-8303-4bb1-8bbb-cc42dacd44d7)\n\nThis stack:\n![image](https://github.com/user-attachments/assets/d6b50b24-c30c-4d23-8b1a-344b3ba65d7a)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135412\nApproved by: https://github.com/oulgen\nghstack dependencies: #135286, #135306, #135377, #135400\n" }, { "commit": "53290ca00b83cf84b1074e0d9e0b7096885a93c4", "tree": "c7554ced5d3b531d697922e71813af86b8785ace", "parents": [ "16f5155992ca291e0fe69d122fae820b6dac057d" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Sat Sep 07 19:54:41 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 18:02:36 2024 +0000" }, "message": "[inductor] Refactor BaseSchedulerNode.__init__ (#135400)\n\nMight be a small compile time improvement since we remove a call to extract_read_writes().\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135400\nApproved by: https://github.com/oulgen\nghstack dependencies: #135286, #135306, #135377\n" }, { "commit": "16f5155992ca291e0fe69d122fae820b6dac057d", "tree": "8204832fcc442b1e2a325950567856f8b29b88f4", "parents": [ "37144be03dc1f95f6e511c328f68367cfc8f94c9" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Sat Sep 07 19:54:40 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 18:02:32 2024 +0000" }, "message": "[inductor] Fast path for extract_read_writes without tracing (#135377)\n\nBefore (bottom of stack):\n![image](https://github.com/user-attachments/assets/13060ff9-b31d-42a9-8e8f-c50b2bf3dc2f)\n\nAfter (this PR):\n![image](https://github.com/user-attachments/assets/7d190821-b614-46b7-9e9e-9087443df654)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135377\nApproved by: https://github.com/oulgen\nghstack dependencies: #135286, #135306\n" }, { "commit": "37144be03dc1f95f6e511c328f68367cfc8f94c9", "tree": "873f8a7fb207b6351d8ba916835e386240267072", "parents": [ "3bdc54ed18824abe93ba7897cd6b9ff91a98c69b" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Sat Sep 07 19:54:40 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 18:02:28 2024 +0000" }, "message": "[inductor] Remove ReadWrites.op_counts (#135306)\n\nThis was (almost) unused.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135306\nApproved by: https://github.com/oulgen\nghstack dependencies: #135286\n" }, { "commit": "3bdc54ed18824abe93ba7897cd6b9ff91a98c69b", "tree": "32b6c1d81051e03a406f9dfc3505b7b809b41f8c", "parents": [ "2196f3247533f2e1b2f205f81ceca9eeb770013e" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Sat Sep 07 19:54:40 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 18:02:24 2024 +0000" }, "message": "[inductor] Refactor LoopBody.memory_usage (#135286)\n\nThis is preparing for some other changes where I speed up extract_read_writes tracing.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135286\nApproved by: https://github.com/oulgen\n" }, { "commit": "2196f3247533f2e1b2f205f81ceca9eeb770013e", "tree": "f0c54eb9bda774f95e760e6d67f8d5949ee01c82", "parents": [ "cfc227ad43c8a39e52463ef06029dddc00919eae" ], "author": { "name": "cyy", "email": "cyyever@outlook.com", "time": "Sun Sep 08 17:18:29 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 17:18:29 2024 +0000" }, "message": "[22/N] Fix clang-tidy warnings in jit (#135319)\n\nFollows #134537\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135319\nApproved by: https://github.com/titaiwangms\n" }, { "commit": "cfc227ad43c8a39e52463ef06029dddc00919eae", "tree": "75c128403de2ed73e3f933cd8cb2c36d39337395", "parents": [ "20cab91a123d74d2c051b80501601ef7a26bc303" ], "author": { "name": "Wanchao Liang", "email": "wanchaol@users.noreply.github.com", "time": "Sun Sep 08 17:08:40 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 17:08:40 2024 +0000" }, "message": "[reland][dtensor] move DTensor to public namespace (#134203)\n\nreland of https://github.com/pytorch/pytorch/pull/133113\n\nI have to create a new PR because the previous reverted PR could not either be rebased, or imported successfully :(\n\n----\n\nMoving DTensor to be in the public namespace, to formally add the documentation page that includes all the public APIs. This includes:\n\n* many path renames and path import fixes\n* a dedicated doc page without too much content yet (adding in the next PRs)\n* To preserve the BC for users still using the torch.distributed._tensor, I added a shim script to redirect old path calls to the new module\n\nThe BC preserving is evidented by the fact that all DTensor tests are still working without changing the public imports. So it\u0027s safe to land the changes\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/134203\nApproved by: https://github.com/tianyu-l\n" }, { "commit": "20cab91a123d74d2c051b80501601ef7a26bc303", "tree": "ae7662613786e03f4016adf1193e1d6b532bc625", "parents": [ "a6fae2e81118cf6f408d2db4ea40e0d7fa145ad6" ], "author": { "name": "Animesh Jain", "email": "anijain@umich.edu", "time": "Thu Sep 05 16:27:22 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 15:11:12 2024 +0000" }, "message": "[dynamo] Remove skip from jit freeze tests (#135281)\n\nFixes https://github.com/pytorch/pytorch/issues/119781\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135281\nApproved by: https://github.com/zou3519\n" }, { "commit": "a6fae2e81118cf6f408d2db4ea40e0d7fa145ad6", "tree": "5bf75ccab9e350683ce4fa16902404347b11ef89", "parents": [ "042f2f7746a064f1527d95d1f1d712b4f0b34186" ], "author": { "name": "CaoE", "email": "e.cao@intel.com", "time": "Fri Sep 06 21:42:54 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 12:32:23 2024 +0000" }, "message": "Use BRGEMM for Half flash attention forward kernel (#131879)\n\nUse oneDNN BRGEMM on packed data to get better performance on the 5th generation of Xeon where Intel® Advanced Matrix Extensions (AMX) will have fp16 support, e.g. amx-fp16.\nMultiple models have achieved acceleration, for instance, FP16 stable diffusion v2.1 has achieved over 50% improvement.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/131879\nApproved by: https://github.com/jgong5, https://github.com/peterbell10\nghstack dependencies: #131878\n" }, { "commit": "042f2f7746a064f1527d95d1f1d712b4f0b34186", "tree": "314b8f5f7c97bf473f15886cfb4d1f321b20ac3d", "parents": [ "fd494dd4266620d0c08ec09f817df55690b5a240" ], "author": { "name": "Justin Chu", "email": "justinchu@microsoft.com", "time": "Sun Sep 08 05:30:34 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 05:30:34 2024 +0000" }, "message": "[ONNX] Re-raise the exception if the dynamic shapes cannot be refined (#135418)\n\nImprove error reporting. Otherwise users will just see not being able to refine shapes most of the time.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135418\nApproved by: https://github.com/titaiwangms\n" }, { "commit": "fd494dd4266620d0c08ec09f817df55690b5a240", "tree": "912ae04cb2bd0ba6c810290b82b7d0f46496ef30", "parents": [ "8334cb2fb91ad16bd37efcb33b8915f6d629eb3f" ], "author": { "name": "Huamin Li", "email": "huaminli@meta.com", "time": "Sun Sep 08 04:16:24 2024 +0000" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 04:16:24 2024 +0000" }, "message": "Change wrapped_linear_prepack and wrapped_quantized_linear_prepacked to private by adding _ as prefix (#135401)\n\nSummary: In https://github.com/pytorch/pytorch/pull/134232, we added two new ops wrapped_linear_prepack and wrapped_quantized_linear_prepacked. From the review comments and offline discussion, we are changing them to private by adding `_` as prefix\n\nDifferential Revision: D62325142\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135401\nApproved by: https://github.com/houseroad\n" }, { "commit": "8334cb2fb91ad16bd37efcb33b8915f6d629eb3f", "tree": "afcb65c706bb67a9fe8a2dd4a05298163ffe8663", "parents": [ "e72ed4717eab2c3d6c54a93537f5b298f33dff21" ], "author": { "name": "Bob Ren", "email": "bobren@fb.com", "time": "Sat Sep 07 16:53:27 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sun Sep 08 02:15:45 2024 +0000" }, "message": "remove commented out breakpoints (#135363)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135363\nApproved by: https://github.com/oulgen\n" }, { "commit": "e72ed4717eab2c3d6c54a93537f5b298f33dff21", "tree": "23a4403ba65e49b83d7cf1c61c5470fe9fb08a02", "parents": [ "3bebc09be9845c0779f190489e8d4caa9e2653c8" ], "author": { "name": "Yanbo Liang", "email": "ybliang8@gmail.com", "time": "Fri Sep 06 22:21:46 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sat Sep 07 19:16:29 2024 +0000" }, "message": "[Dynamo] Fix Huggingface PretrainedConfig get non const attr (#135413)\n\nFixes #135329\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135413\nApproved by: https://github.com/anijain2305\n" }, { "commit": "3bebc09be9845c0779f190489e8d4caa9e2653c8", "tree": "9bf9b2a2237a7a03c5b0b50180f8c5d17f037cd7", "parents": [ "a2db22e6bb2889af6672e3c2ebfd9385720d84fa" ], "author": { "name": "drisspg", "email": "drisspguessous@gmail.com", "time": "Fri Sep 06 15:35:07 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sat Sep 07 16:33:41 2024 +0000" }, "message": "[FlexAttention] Align the matmul tensorcore usage (#135168)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135168\nApproved by: https://github.com/Chillee\n" }, { "commit": "a2db22e6bb2889af6672e3c2ebfd9385720d84fa", "tree": "f420995d93fc9eae704dbee85aef7043a6d49fd7", "parents": [ "eac5e1254883e693b028db66e7fc90b3679acbd0" ], "author": { "name": "Sam Larsen", "email": "slarsen@meta.com", "time": "Fri Sep 06 11:40:31 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sat Sep 07 16:33:37 2024 +0000" }, "message": "[inductor] Catch BrokenProcessPool and print a more helpful message. (#135120)\n\nSummary: BrokenProcessPool means a parallel-compile subprocess exited, which we never expect. It\u0027s likely due to a crash, so print a more meaningful error message and instructions that it\u0027s probably easier to debug by turning off parallel compile. Output looks like:\n```\n...\n File \"/data/users/slarsen/pytorch/torch/_inductor/runtime/compile_tasks.py\", line 45, in _reload_python_module\n exec(code, mod.__dict__, mod.__dict__)\n File \"/tmp/torchinductor_slarsen/4q/c4qw7xk5lbb7whg5txnk4hwbc7z6kepak3o666tr3d64gcad5r5b.py\", line 815, in \u003cmodule\u003e\n async_compile.wait(globals())\n File \"/data/users/slarsen/pytorch/torch/_inductor/async_compile.py\", line 265, in wait\n raise RuntimeError(\nRuntimeError: A compilation subprocess exited unexpectedly. This is likely due to a crash. To facilitate debugging, you can re-run with TORCHINDUCTOR_COMPILE_THREADS\u003d1 to cause compilation to occur in the main process.\n```\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135120\nApproved by: https://github.com/Chillee\n" }, { "commit": "eac5e1254883e693b028db66e7fc90b3679acbd0", "tree": "023321e001fe38a44b4e48c2cea3fe0295c9e634", "parents": [ "18479c5f702ec4082714714f04e67bdefd056e78" ], "author": { "name": "Jason Ansel", "email": "jansel@meta.com", "time": "Fri Sep 06 21:17:18 2024 -0700" }, "committer": { "name": "PyTorch MergeBot", "email": "pytorchmergebot@users.noreply.github.com", "time": "Sat Sep 07 16:29:15 2024 +0000" }, "message": "[inductor] Move LoopBody to its own file (#135257)\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/135257\nApproved by: https://github.com/oulgen\n" } ], "next": "18479c5f702ec4082714714f04e67bdefd056e78" }