{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":730932260,"defaultBranch":"main","name":"torchtitan","ownerLogin":"pytorch","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2023-12-13T01:51:37.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/21003710?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1716948940.0","currentOid":""},"activityList":{"items":[{"before":"54aecd22702e5b1f5b10910de01c270ef3d7702a","after":"0f928ee3b4057f2b311f8fc0f6750ba7edc961f9","ref":"refs/heads/gh/tianyu-l/12/orig","pushedAt":"2024-05-29T03:46:54.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"only produce tensorboard logs on rank 0 by default\n\nghstack-source-id: 8d4a50e453d0be2b4a4400ac09a1a793ce8726e5\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/339","shortMessageHtmlLink":"only produce tensorboard logs on rank 0 by default"}},{"before":"db6f334c1258ead929826d5d4056bfdb809b0a6f","after":"5d79a23dabd9c34762c50cadf6147d9902297e6c","ref":"refs/heads/gh/tianyu-l/12/head","pushedAt":"2024-05-29T03:46:52.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"only produce tensorboard logs on rank 0 by default\""}},{"before":"4ede4f94231673ced7b05d56f65aa8e63548e76d","after":"95741b9f010a99e0f68514d5bd4a10fc25263e58","ref":"refs/heads/gh/tianyu-l/12/base","pushedAt":"2024-05-29T03:46:51.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update base for Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"only produce tensorboard logs on rank 0 by…"}},{"before":"d004ea8e74744f97fa426ca760d5d2efe6cec3d6","after":"309e5acf752635fe8b81a9d5eb06f5ea2a9d4477","ref":"refs/heads/gh/wconstab/30/orig","pushedAt":"2024-05-29T03:15:29.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Fix 1D PP tracer test, add 2D test\n\nforgot to enable tracer for tracer test in the last PR\n\nghstack-source-id: 6eff83d7fe5af576dc6da0dcdae5bc51b4ac8ec4\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/362","shortMessageHtmlLink":"Fix 1D PP tracer test, add 2D test"}},{"before":"c142d3a06591f1bbb08375a250a86a985ba69b3d","after":"98b6432514c3faaa98e9306529508907c64f15a0","ref":"refs/heads/gh/kwen2501/3/orig","pushedAt":"2024-05-29T03:15:29.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Use general way to access and update submodules\n\nghstack-source-id: 83e0f747ab4c68b639a4a9d7100d09d6bca653ed\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/371","shortMessageHtmlLink":"Use general way to access and update submodules"}},{"before":"cf702832a5f7cf54b311980e30e5141322bff557","after":"435b3ca4cdc8322df1d23c7eef15784b46a01906","ref":"refs/heads/gh/wconstab/30/head","pushedAt":"2024-05-29T03:15:27.000Z","pushType":"push","commitsCount":196,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update on \"Fix 1D PP tracer test, add 2D test\"\n\n\nforgot to enable tracer for tracer test in the last PR\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"Fix 1D PP tracer test, add 2D test\""}},{"before":"f85e0fbfd881cde26d9923551653d6cf2591a53b","after":"7acdbf88068f404a64b9b437f09be57f979cce4f","ref":"refs/heads/gh/kwen2501/3/head","pushedAt":"2024-05-29T03:15:27.000Z","pushType":"push","commitsCount":196,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update on \"Use general way to access and update submodules\"\n\n\r\nThis PR fixes the issue mentioned [here](https://github.com/pytorch/pytorch/pull/126653#issuecomment-2130504712):\r\n\"Module object has no attributed items.\"\r\n\r\nThe reason is, a split `ModuleDict` is no longer a `ModuleDict`. (Future support is not guaranteed.)\r\n\r\nIt would be more generally applicable if we use `named_children()` and `register_module()` to access and update submodules.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"Use general way to access and update submodules\""}},{"before":"cf702832a5f7cf54b311980e30e5141322bff557","after":"d92eea9529f276bc43070347af45018846b79e07","ref":"refs/heads/gh/kwen2501/3/base","pushedAt":"2024-05-29T03:15:25.000Z","pushType":"push","commitsCount":195,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update base for Update on \"Use general way to access and update submodules\"\n\n\r\nThis PR fixes the issue mentioned [here](https://github.com/pytorch/pytorch/pull/126653#issuecomment-2130504712):\r\n\"Module object has no attributed items.\"\r\n\r\nThe reason is, a split `ModuleDict` is no longer a `ModuleDict`. (Future support is not guaranteed.)\r\n\r\nIt would be more generally applicable if we use `named_children()` and `register_module()` to access and update submodules.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"Use general way to access and update submo…"}},{"before":"f1ee7cb18dcc5fead1ad767f4b15322fc2bf44d8","after":"e76e0f72b15412c1d9760bf2ee290e784c9b136d","ref":"refs/heads/gh/wconstab/30/base","pushedAt":"2024-05-29T03:15:25.000Z","pushType":"push","commitsCount":195,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update base for Update on \"Fix 1D PP tracer test, add 2D test\"\n\n\nforgot to enable tracer for tracer test in the last PR\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"Fix 1D PP tracer test, add 2D test\""}},{"before":"61e0687161646df27ff5f577fc252971fcaf1d29","after":"c142d3a06591f1bbb08375a250a86a985ba69b3d","ref":"refs/heads/gh/kwen2501/3/orig","pushedAt":"2024-05-29T02:17:02.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Use general way to access and update submodules\n\nghstack-source-id: 6eff83d7fe5af576dc6da0dcdae5bc51b4ac8ec4\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/371","shortMessageHtmlLink":"Use general way to access and update submodules"}},{"before":"eb2aa68b06cc5dfecfc1956de56c230a2bdf1715","after":"f85e0fbfd881cde26d9923551653d6cf2591a53b","ref":"refs/heads/gh/kwen2501/3/head","pushedAt":"2024-05-29T02:17:00.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update on \"Use general way to access and update submodules\"\n\n\n\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"Use general way to access and update submodules\""}},{"before":null,"after":"61e0687161646df27ff5f577fc252971fcaf1d29","ref":"refs/heads/gh/kwen2501/3/orig","pushedAt":"2024-05-29T02:15:40.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Use general way to access and update submodules\n\nghstack-source-id: 9372cca2b62078993fa7cd8f28d3b5bb710b406b\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/371","shortMessageHtmlLink":"Use general way to access and update submodules"}},{"before":null,"after":"cf702832a5f7cf54b311980e30e5141322bff557","ref":"refs/heads/gh/kwen2501/3/base","pushedAt":"2024-05-29T02:15:35.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Update\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update"}},{"before":null,"after":"eb2aa68b06cc5dfecfc1956de56c230a2bdf1715","ref":"refs/heads/gh/kwen2501/3/head","pushedAt":"2024-05-29T02:15:35.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Use general way to access and update submodules\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Use general way to access and update submodules"}},{"before":null,"after":"8d67f222abfd9afc4dc212357779d5283c9b0983","ref":"refs/heads/gh/kwen2501/2/base","pushedAt":"2024-05-29T02:12:37.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Add --test option to specify test to run\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Add --test option to specify test to run"}},{"before":null,"after":"cfe44da21997110a6a659342aecbc2e9e0a6aa60","ref":"refs/heads/gh/kwen2501/2/head","pushedAt":"2024-05-29T02:12:37.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Lint\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Lint"}},{"before":null,"after":"1ceaa4e2adc8ef5a0864f99e126e4ab18cd7db8f","ref":"refs/heads/gh/kwen2501/1/base","pushedAt":"2024-05-29T02:12:34.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Add test for PP tracer frontend\n\n- switch to using public PipelineStage API\n- clean up some asserts in tracer codepath\n\nghstack-source-id: 2d069b7d45c4f3c788dec8fc85d8a7e83e463fcd\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/357","shortMessageHtmlLink":"Add test for PP tracer frontend"}},{"before":null,"after":"8d67f222abfd9afc4dc212357779d5283c9b0983","ref":"refs/heads/gh/kwen2501/1/head","pushedAt":"2024-05-29T02:12:34.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Add --test option to specify test to run\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Add --test option to specify test to run"}},{"before":"98c17df5e84a6ebaac12db3b40e3465bf4f75c3b","after":"54aecd22702e5b1f5b10910de01c270ef3d7702a","ref":"refs/heads/gh/tianyu-l/12/orig","pushedAt":"2024-05-29T01:30:23.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"only produce tensorboard logs on rank 0 by default\n\nghstack-source-id: c6cf5ef43918478b27d65944ec1c217cf2794fe2\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/339","shortMessageHtmlLink":"only produce tensorboard logs on rank 0 by default"}},{"before":"9c27dfb9d98ad3257880c43a4c029af430b7fd66","after":"db6f334c1258ead929826d5d4056bfdb809b0a6f","ref":"refs/heads/gh/tianyu-l/12/head","pushedAt":"2024-05-29T01:30:22.000Z","pushType":"push","commitsCount":2,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"only produce tensorboard logs on rank 0 by default\""}},{"before":"6ffcf78727e8a35b6f025d3f7bcc2209b50d514e","after":"4ede4f94231673ced7b05d56f65aa8e63548e76d","ref":"refs/heads/gh/tianyu-l/12/base","pushedAt":"2024-05-29T01:30:20.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update base for Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"only produce tensorboard logs on rank 0 by…"}},{"before":"2f712ccd8c6e62965777c9a5f82d461a43e8830c","after":"d4b10b81de71ef95bcf4761dec860e413b0fd2a8","ref":"refs/heads/test_id","pushedAt":"2024-05-29T01:29:19.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Lint","shortMessageHtmlLink":"Lint"}},{"before":null,"after":"2f712ccd8c6e62965777c9a5f82d461a43e8830c","ref":"refs/heads/test_id","pushedAt":"2024-05-29T01:22:41.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"kwen2501","name":"Ke Wen","path":"/kwen2501","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6676466?s=80&v=4"},"commit":{"message":"Add --test option to specify test to run","shortMessageHtmlLink":"Add --test option to specify test to run"}},{"before":null,"after":"7ce802e0bd76d308a37c2286bda62f999831713b","ref":"refs/heads/gh/wconstab/31/orig","pushedAt":"2024-05-29T00:38:28.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"wconstab","name":"Will Constable","path":"/wconstab","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4984825?s=80&v=4"},"commit":{"message":"test changes\n\nghstack-source-id: 5c2a30d588238f1d1d69356874fa93ed6c2e7cba\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/367","shortMessageHtmlLink":"test changes"}},{"before":null,"after":"49763615ea77b9a5e42726e0cdf3b376ad876629","ref":"refs/heads/gh/wconstab/31/head","pushedAt":"2024-05-29T00:38:23.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"wconstab","name":"Will Constable","path":"/wconstab","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4984825?s=80&v=4"},"commit":{"message":"Update\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update"}},{"before":null,"after":"537b913225826806223b567f53406819d7b1832d","ref":"refs/heads/gh/wconstab/31/base","pushedAt":"2024-05-29T00:38:23.000Z","pushType":"branch_creation","commitsCount":0,"pusher":{"login":"wconstab","name":"Will Constable","path":"/wconstab","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/4984825?s=80&v=4"},"commit":{"message":"Update (base update)\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update (base update)"}},{"before":"2762a0a22461b6301ef000b5559151a3d586f5cd","after":"bf50097b7a94894f5bb74eee62c5b6217af160a3","ref":"refs/heads/compile_norm","pushedAt":"2024-05-28T22:30:30.000Z","pushType":"push","commitsCount":17,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"repro compiled rmsnorm error","shortMessageHtmlLink":"repro compiled rmsnorm error"}},{"before":"dbcc6dccffad3d85a7099ce18e3ead40c6b83a80","after":"98c17df5e84a6ebaac12db3b40e3465bf4f75c3b","ref":"refs/heads/gh/tianyu-l/12/orig","pushedAt":"2024-05-28T21:12:53.000Z","pushType":"force_push","commitsCount":0,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"only produce tensorboard logs on rank 0 by default\n\nghstack-source-id: 1fbc146696046326bff72cfeb192625ccfda055e\nPull Request resolved: https://github.com/pytorch/torchtitan/pull/339","shortMessageHtmlLink":"only produce tensorboard logs on rank 0 by default"}},{"before":"7c88f8bf1368a131b14c59e47705bcba8b6e2101","after":"9c27dfb9d98ad3257880c43a4c029af430b7fd66","ref":"refs/heads/gh/tianyu-l/12/head","pushedAt":"2024-05-28T21:12:52.000Z","pushType":"push","commitsCount":4,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update on \"only produce tensorboard logs on rank 0 by default\""}},{"before":"f31899115563d427ae3e2b5d48c1c446f1bd972a","after":"6ffcf78727e8a35b6f025d3f7bcc2209b50d514e","ref":"refs/heads/gh/tianyu-l/12/base","pushedAt":"2024-05-28T21:12:50.000Z","pushType":"push","commitsCount":3,"pusher":{"login":"tianyu-l","name":null,"path":"/tianyu-l","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/150487191?s=80&v=4"},"commit":{"message":"Update base for Update on \"only produce tensorboard logs on rank 0 by default\"\n\n\r\n1. For tensorboard metrics, we mostly care about loss, memory, wps/mfu. Loss is all-reduced so will be the same on all ranks; other metrics are likely to be very similar among all ranks. So by default it suffices to only do tb logging on rank 0 -- the straggler effect should be small for tb writes. User could always toggle on all-rank logging for debugging purposes.\r\n\r\n2. Remove `torch` dependency in `requirements.txt` as it cannot work alone / is not used anyways. Currently we are suggesting users to install latest nightly in README, and do so in all the CI tests.\n\n[ghstack-poisoned]","shortMessageHtmlLink":"Update base for Update on \"only produce tensorboard logs on rank 0 by…"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEVn4IVQA","startCursor":null,"endCursor":null}},"title":"Activity · pytorch/torchtitan"}