{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":65600975,"defaultBranch":"main","name":"pytorch","ownerLogin":"pytorch","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2016-08-13T05:26:41.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/21003710?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1718147059.0","currentOid":""},"activityList":{"items":[{"before":"840e45b57997602f698db8dbc9c82cc286b1f3cc","after":null,"ref":"refs/tags/ciflow/inductor/127431","pushedAt":"2024-06-11T22:57:03.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"647b448f2ff1327ba141d461f2b9f0b2212f50ba","after":"9f1ffaffeb049c8f0fadc0d80b606e0befefc494","ref":"refs/heads/gh/tugsbayasgalan/220/orig","pushedAt":"2024-06-11T22:56:59.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"tugsbayasgalan","name":"Tugsbayasgalan Manlaibaatar","path":"/tugsbayasgalan","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/16603271?s=80&v=4"},"commit":{"message":"[DRAFT] nested tensor subclass support\n\nghstack-source-id: f1f21170aab1eeaacf0ad03a245916c3061c2031\nPull Request resolved: https://github.com/pytorch/pytorch/pull/127431","shortMessageHtmlLink":"[DRAFT] nested tensor subclass support"}},{"before":"840e45b57997602f698db8dbc9c82cc286b1f3cc","after":"aa5a8f6de47cfc6938210fb1c85e283aa9a24ce4","ref":"refs/heads/gh/tugsbayasgalan/220/head","pushedAt":"2024-06-11T22:56:56.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"tugsbayasgalan","name":"Tugsbayasgalan Manlaibaatar","path":"/tugsbayasgalan","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/16603271?s=80&v=4"},"commit":{"message":"Update on \"[DRAFT] nested tensor subclass support\"\n\n\r\nWhen we have nested tensor subclasses, we need to recurse down to access the underlying real tensor and wrap it in FakeTensor and recursively build back up the nested tensor subclasses. I am not sure if I am passing around the SymbolicContext correctly?\r\n\r\n\r\ncc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"[DRAFT] nested tensor subclass support\""}},{"before":"87282525b5b73f4bce672659509976f4656e8415","after":"85da42d29b0a2a53c4354826b60e0710402663f0","ref":"refs/heads/gh/tugsbayasgalan/220/base","pushedAt":"2024-06-11T22:56:52.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"tugsbayasgalan","name":"Tugsbayasgalan Manlaibaatar","path":"/tugsbayasgalan","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/16603271?s=80&v=4"},"commit":{"message":"Update base for Update on \"[DRAFT] nested tensor subclass support\"\n\n\r\nWhen we have nested tensor subclasses, we need to recurse down to access the underlying real tensor and wrap it in FakeTensor and recursively build back up the nested tensor subclasses. I am not sure if I am passing around the SymbolicContext correctly?\r\n\r\n\r\ncc voznesenskym penguinwu EikanWang jgong5 Guobing-Chen XiaobingSuper zhuhaozhe blzheng wenzhe-nrv jiayisunx chenyang78 kadeng chauhang\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"[DRAFT] nested tensor subclass support\""}},{"before":"698b101dc38edfc1efa1cfd8c461a57ea7526715","after":null,"ref":"refs/tags/ciflow/inductor/128054","pushedAt":"2024-06-11T22:55:13.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"dad715460ac634239bf3a68bc50bfee48afb1be8","after":"952422e60ede96106f9e6c0c8729ee11c94d37dc","ref":"refs/heads/gh/isuruf/56/orig","pushedAt":"2024-06-11T22:55:10.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"isuruf","name":"Isuru Fernando","path":"/isuruf","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5234427?s=80&v=4"},"commit":{"message":"Improve codegen for ops.masked in triton\n\nghstack-source-id: 0de345146a15cab17e7e60842ed986fa402332b3\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128054","shortMessageHtmlLink":"Improve codegen for ops.masked in triton"}},{"before":"698b101dc38edfc1efa1cfd8c461a57ea7526715","after":"a48b7bbe07e0ca6c9706b20965ba8f20408a74f3","ref":"refs/heads/gh/isuruf/56/head","pushedAt":"2024-06-11T22:55:06.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"isuruf","name":"Isuru Fernando","path":"/isuruf","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/5234427?s=80&v=4"},"commit":{"message":"Update\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update"}},{"before":"710b7572217638574c73428a6bfb83e2aa624825","after":null,"ref":"refs/tags/ciflow/periodic/128004","pushedAt":"2024-06-11T22:52:53.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"710b7572217638574c73428a6bfb83e2aa624825","after":null,"ref":"refs/tags/ciflow/trunk/128004","pushedAt":"2024-06-11T22:52:51.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"7da694206698783f2e33bc149078b06b3f7378b9","after":"2276d4e9d08c6f8ae3a9a0640cfecc2be5e140dd","ref":"refs/heads/gh/sinhaanhsul/15/orig","pushedAt":"2024-06-11T22:49:38.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"sinhaanshul","name":null,"path":"/sinhaanshul","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/50644008?s=80&v=4"},"commit":{"message":"[debug][be] improving readability of comm_mode.py and comm_mode_features_example.py\n\nghstack-source-id: 01fbb146f58da79f07031014a20b968614bafb32\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128451","shortMessageHtmlLink":"[debug][be] improving readability of comm_mode.py and comm_mode_featu…"}},{"before":"38f85ff6d674ec961ea46b40e2fe0d318dbf9f70","after":"6fef83afc4f1e2081f270cab8209d80aaf0c46e8","ref":"refs/heads/gh/sinhaanhsul/14/orig","pushedAt":"2024-06-11T22:49:38.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"sinhaanshul","name":null,"path":"/sinhaanshul","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/50644008?s=80&v=4"},"commit":{"message":"[dtensor][debug] add module level tracing and readable display\n\nghstack-source-id: 3e9dbb6a4d51b58f3d4f9973df3823df5fb48cd8\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128369","shortMessageHtmlLink":"[dtensor][debug] add module level tracing and readable display"}},{"before":null,"after":"ee08bab7951b52f041c8e55c0e43e3e74c34cab1","ref":"refs/heads/gh/sinhaanhsul/16/orig","pushedAt":"2024-06-11T22:49:38.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"sinhaanshul","name":null,"path":"/sinhaanshul","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/50644008?s=80&v=4"},"commit":{"message":"[dtensor][example] added MLPStacked example for printing sharding\n\nghstack-source-id: e41551565be62a68f5751c3e2bc4c3e869252082\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128461","shortMessageHtmlLink":"[dtensor][example] added MLPStacked example for printing sharding"}},{"before":null,"after":"0f977c157fe9ebc821b2a6606685a8b016d28833","ref":"refs/heads/gh/sinhaanhsul/16/head","pushedAt":"2024-06-11T22:49:31.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"sinhaanshul","name":null,"path":"/sinhaanshul","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/50644008?s=80&v=4"},"commit":{"message":"Update\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update"}},{"before":null,"after":"b8d0b6715d53eb833d033924a74bf209b3044536","ref":"refs/heads/gh/sinhaanhsul/16/base","pushedAt":"2024-06-11T22:49:31.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"sinhaanshul","name":null,"path":"/sinhaanshul","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/50644008?s=80&v=4"},"commit":{"message":"Update (base update)\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update (base update)"}},{"before":"e29ea1a71fc7ea20447d168cde6a8c3a899a2121","after":"8a09940a543d4c2fd23a5c78edbf1ac24d481b45","ref":"refs/heads/fbcode/pytorch-stable-prototype","pushedAt":"2024-06-11T22:43:07.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"bigfootjon","name":"Jon Janzen","path":"/bigfootjon","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6403568?s=80&v=4"},"commit":{"message":"[inductor] fix compile time regression by caching get_gpu_type (#128363)\n\nWe observed signficant compile time regression in torchtitan when turning\non 2D parallel + torch.compile recently. So I decided to get a deeper\nunderstanding why.\n\nIt turns out this is affecting **all the trainings** that have functional collectives\ncaptured in the graph, not only 2D parallel (2D parallel was just the\njob that happen to have collectives captured in the TP region).\n\nThe root cause is because when doing inductor lowering, we are calling\nthe comm analysis pass to get a estimated collective time for each\ncollective node in the graph, for each call to check the collective\nnode, we are calling `get_gpu_type()`, which under the hood calls a\n`torch.utils.collect_env.run` to get the GPU info. However, this call is\nsuper expensive! The reason is that this call effectively spawns a new\nprocess and call `nvidia-smi` to get the GPU info, so the cost is **linear**\nto the number of collective nodes in the graph.\n\nsee https://github.com/pytorch/pytorch/blob/main/torch/utils/collect_env.py#L75\n\nThe fix is to add a lru cache to the function, so that we only call this\nonce and reuse the cached results afterwards\n\ntorchtitan benchmark shows:\n* before this fix: 2D parallel + fp8 compile time: 6min +\n* after this fix: 2D parallel + fp8 compile time: 2min 48s (more than 100% improvement)\n\nThere're more room to improve the compile time, but this PR is trying to fix the biggest regression I found so far.\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128363\nApproved by: https://github.com/yf225","shortMessageHtmlLink":"[inductor] fix compile time regression by caching get_gpu_type (<a class=\"issue-link js-issue-link\" data-error-text=\"Failed to load title\" data-id=\"2345010433\" data-permission-text=\"Title is private\" data-url=\"https://github.com/pytorch/pytorch/issues/128363\" data-hovercard-type=\"pull_request\" data-hovercard-url=\"/pytorch/pytorch/pull/128363/hovercard\" href=\"https://github.com/pytorch/pytorch/pull/128363\">#128363</a>)"}},{"before":"447173198b9ff16a908c402b0a077da7489bbb81","after":"94fea82d6646c3f55f80d6a4e84a4104e18387f4","ref":"refs/heads/main","pushedAt":"2024-06-11T22:42:36.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchmergebot","name":null,"path":"/pytorchmergebot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/97764156?s=80&v=4"},"commit":{"message":"init sub comment (#128082)\n\nFixes #127905\n\n### Description\n\nAdd docstring to torch/onnx/symbolic_opset9.py:sigmoid function\n\n### Checklist\n- [x] The issue that is being fixed is referred in the description\n- [x] Only one issue is addressed in this pull request\n- [x] Labels from the issue that this PR is fixing are added to this pull request\n- [x] No unnecessary issues are included into this pull request\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128082\nApproved by: https://github.com/titaiwangms","shortMessageHtmlLink":"init sub comment (<a class=\"issue-link js-issue-link\" data-error-text=\"Failed to load title\" data-id=\"2336987377\" data-permission-text=\"Title is private\" data-url=\"https://github.com/pytorch/pytorch/issues/128082\" data-hovercard-type=\"pull_request\" data-hovercard-url=\"/pytorch/pytorch/pull/128082/hovercard\" href=\"https://github.com/pytorch/pytorch/pull/128082\">#128082</a>)"}},{"before":"f881fe428acf8006baa28c9b725e3bf2cee93a82","after":null,"ref":"refs/tags/ciflow/trunk/128139","pushedAt":"2024-06-11T22:42:18.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"b79d056e76ac2644e134c016053ec15c119b53f8","after":"447173198b9ff16a908c402b0a077da7489bbb81","ref":"refs/heads/main","pushedAt":"2024-06-11T22:42:13.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchmergebot","name":null,"path":"/pytorchmergebot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/97764156?s=80&v=4"},"commit":{"message":"Add docstring for the torch.fx.operator_schemas.create_type_hint func… (#128139)\n\nFixes: #127916\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128139\nApproved by: https://github.com/SherlockNoMad","shortMessageHtmlLink":"Add docstring for the torch.fx.operator_schemas.create_type_hint func… ("}},{"before":"10a7e6051c4e8993ec796f7d2c35082f1592a390","after":null,"ref":"refs/tags/ciflow/mps/128421","pushedAt":"2024-06-11T22:37:12.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"e1f738fd7f96c9323d33fbf509f97ae5e1b7bf92","after":"4feade294ebff142b7cc03f0f2cf80a8ac717c20","ref":"refs/heads/gh/malfet/3/orig","pushedAt":"2024-06-11T22:37:10.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"malfet","name":"Nikita Shulga","path":"/malfet","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2453524?s=80&v=4"},"commit":{"message":"[MPS] Add Metal implementation of exp op\n\nTo improve accuracy, use only `exp` from `precise::` namespace\n\nghstack-source-id: 05135fad3f80582cfcca5ac4f3c46c8cf835fff8\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128421","shortMessageHtmlLink":"[MPS] Add Metal implementation of exp op"}},{"before":"10a7e6051c4e8993ec796f7d2c35082f1592a390","after":"214b31a6fa4d79921a4d36eef03a2159043ebd8a","ref":"refs/heads/gh/malfet/3/head","pushedAt":"2024-06-11T22:37:07.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"malfet","name":"Nikita Shulga","path":"/malfet","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/2453524?s=80&v=4"},"commit":{"message":"Update on \"[MPS] Add Metal implementation of exp op\"\n\n\r\nTo improve accuracy, use `precise::exp()` (and `precise::sin()`/`precise::cos()` for complex flavor)\r\n\r\nFix bug in non-contiguous tensors handling\r\n\r\nFixes https://github.com/pytorch/pytorch/issues/84936 \n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"[MPS] Add Metal implementation of exp op\""}},{"before":"125a48b0f1df20623545ee10586e544e884b40ab","after":"494abb77b106d5ee77d040f95c447bb508174e8c","ref":"refs/heads/gh/jamesperng/2/orig","pushedAt":"2024-06-11T22:36:30.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"jamesperng","name":"James Perng","path":"/jamesperng","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/19395713?s=80&v=4"},"commit":{"message":"[pytorch] add folly:Indestructible\n\nPull Request resolved: https://github.com/pytorch/pytorch/pull/128425\n\n\n\n\n@imported-using-ghimport\n\nDifferential Revision: [D58432297](https://our.internmc.facebook.com/intern/diff/D58432297/)\nghstack-source-id: 229838786","shortMessageHtmlLink":"[pytorch] add folly:Indestructible"}},{"before":"07c968c304ba54f96a911e6a695fa344e873f089","after":"10b062574ec801d305d3b3eeb9f6b8b42f4d3f31","ref":"refs/heads/gh/jamesperng/2/head","pushedAt":"2024-06-11T22:36:29.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"jamesperng","name":"James Perng","path":"/jamesperng","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/19395713?s=80&v=4"},"commit":{"message":"Update on \"[pytorch] add folly:Indestructible\"\n\n\nDifferential Revision: [D58432297](https://our.internmc.facebook.com/intern/diff/D58432297)\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"[pytorch] add folly:Indestructible\""}},{"before":"3d4cbc27dcc0da51ce5fd88d8502961d979f8c5c","after":"71034bbfdd8c84a22443f54b6f50c14b8be15024","ref":"refs/heads/gh/jamesperng/2/base","pushedAt":"2024-06-11T22:36:27.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"jamesperng","name":"James Perng","path":"/jamesperng","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/19395713?s=80&v=4"},"commit":{"message":"Update base for Update on \"[pytorch] add folly:Indestructible\"\n\n\nDifferential Revision: [D58432297](https://our.internmc.facebook.com/intern/diff/D58432297)\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"[pytorch] add folly:Indestructible\""}},{"before":"92ec2cff3679b778e34c11e1341a24275d2aac6f","after":null,"ref":"refs/tags/ciflow/trunk/128260","pushedAt":"2024-06-11T22:32:17.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"2d9424a0c0a1dffcf0ff3f47c2702b2fa91a5871","after":null,"ref":"refs/tags/ciflow/inductor/126578","pushedAt":"2024-06-11T22:32:15.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"287482cd133bf729242574ce46e6d01f768b6b07","after":null,"ref":"refs/tags/ciflow/inductor/128428","pushedAt":"2024-06-11T22:32:15.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"aed81f2c4f2293655dcc8c24da6d8f71a3e395cb","after":null,"ref":"refs/tags/ciflow/inductor/128453","pushedAt":"2024-06-11T22:32:15.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"3039c055bd3fbec2cf65a1afd2d0a87ac3384273","after":null,"ref":"refs/tags/ciflow/inductor/128315","pushedAt":"2024-06-11T22:32:15.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}},{"before":"8b0284cf554ebbf3eee865cf0fbf64f18da2cf94","after":null,"ref":"refs/tags/ciflow/inductor/128148","pushedAt":"2024-06-11T22:32:14.000Z","pushType":"branch_deletion","commitsCount":0,"pusher":{"login":"pytorch-bot[bot]","name":null,"path":"/apps/pytorch-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/40112?s=80&v=4"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEYrlmRgA","startCursor":null,"endCursor":null}},"title":"Activity · pytorch/pytorch"}