{"payload":{"feedbackUrl":"https://github.com/orgs/community/discussions/53140","repo":{"id":150154628,"defaultBranch":"main","name":"FBGEMM","ownerLogin":"pytorch","currentUserCanPush":false,"isFork":false,"isEmpty":false,"createdAt":"2018-09-24T19:07:42.000Z","ownerAvatar":"https://avatars.githubusercontent.com/u/21003710?v=4","public":true,"private":false,"isOrgOwned":true},"refInfo":{"name":"","listCacheKey":"v0:1716168607.0","currentOid":""},"activityList":{"items":[{"before":"071ac24444bd3767114e792a73bd331c6b31d028","after":"e935442e2edfda8980d68104b1fe1fdeaa249d8b","ref":"refs/heads/gh-pages","pushedAt":"2024-05-30T05:19:20.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@2dd8b6343a153519d8edf7eec9e3c6b650f4e164 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 2dd8b63 ๐Ÿš€"}},{"before":"35fa7bedc1f8d1cbea00773859c813a34d7ccbd1","after":"2dd8b6343a153519d8edf7eec9e3c6b650f4e164","ref":"refs/heads/main","pushedAt":"2024-05-30T05:09:09.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Skip multi_gpu_car_test on OSS (#2644)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2644\n\nSkip multi_gpu_car_test on OSS as it requires multiple GPUs.\n\nReviewed By: q10\n\nDifferential Revision: D57949245\n\nfbshipit-source-id: 4fb753cecd7b743b0225ff7defbed8efbf77d00d","shortMessageHtmlLink":"Skip multi_gpu_car_test on OSS (#2644)"}},{"before":"df4de8f2dce88c2c6b4c07718660b260a0217519","after":"071ac24444bd3767114e792a73bd331c6b31d028","ref":"refs/heads/gh-pages","pushedAt":"2024-05-29T17:14:16.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@35fa7bedc1f8d1cbea00773859c813a34d7ccbd1 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 35fa7be ๐Ÿš€"}},{"before":"7b39db296cfa3521c890a2dd95972ec5d17f4a6e","after":"35fa7bedc1f8d1cbea00773859c813a34d7ccbd1","ref":"refs/heads/main","pushedAt":"2024-05-29T17:04:25.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Fix Torch Compile with FP8 Quantization (#2637)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2637\n\nFixes a few incompatibilities between torch compile and fp8 quantization. Outputs look correct but we unfortunately still dont see any speed up. At least this allows further performance analysis.\n\nReviewed By: jianyuh\n\nDifferential Revision: D57885347\n\nfbshipit-source-id: dce775069bde1a8d1891cb52f9813eb54d0d162a","shortMessageHtmlLink":"Fix Torch Compile with FP8 Quantization (#2637)"}},{"before":"6d797095e2054eb35f688df54b9ed4e1348f2c0a","after":"fc0e96b662cd1f5849f90ba79557f01e3fb15a3c","ref":"refs/heads/nightly","pushedAt":"2024-05-29T11:35:02.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-29 nightly release (7b39db296cfa3521c890a2dd95972ec5d17f4a6e)","shortMessageHtmlLink":"2024-05-29 nightly release (7b39db2)"}},{"before":"f2788320ca82b17ab67ec14789eca4fd1d657702","after":"df4de8f2dce88c2c6b4c07718660b260a0217519","ref":"refs/heads/gh-pages","pushedAt":"2024-05-29T03:26:53.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@7b39db296cfa3521c890a2dd95972ec5d17f4a6e ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 7b39db2 ๐Ÿš€"}},{"before":"e80ee7f245935d21939013e81599834ce06350b6","after":"7b39db296cfa3521c890a2dd95972ec5d17f4a6e","ref":"refs/heads/main","pushedAt":"2024-05-29T03:17:28.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"add dense TBE to template and enable VBE support (#2620)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2620\n\n- make the dense TBE headers into a template\n- add VBE options to the dense TBE header\n\nReviewed By: sryap\n\nDifferential Revision: D57017981\n\nfbshipit-source-id: ceb4e42ac8e7494761accde2d7f2103efbc10bc4","shortMessageHtmlLink":"add dense TBE to template and enable VBE support (#2620)"}},{"before":"79ff8396fb5c9995ef271db520675d2aa5b44866","after":"f2788320ca82b17ab67ec14789eca4fd1d657702","ref":"refs/heads/gh-pages","pushedAt":"2024-05-29T00:47:34.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@e80ee7f245935d21939013e81599834ce06350b6 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ e80ee7f ๐Ÿš€"}},{"before":"b633904f888fc02f34a2634aa4de265cfe4af5f0","after":"e80ee7f245935d21939013e81599834ce06350b6","ref":"refs/heads/main","pushedAt":"2024-05-29T00:37:15.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Fix import for codegen (#2636)\n\nSummary:\n- Fix import for codegen\n\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2636\n\nReviewed By: sryap\n\nDifferential Revision: D57880829\n\nPulled By: q10\n\nfbshipit-source-id: 518d51c51cf5ded33aef1204b920f5c92b69c30c","shortMessageHtmlLink":"Fix import for codegen (#2636)"}},{"before":"5bd335fd0f8f0fce831c7fb49f96d970ec95c96d","after":"b633904f888fc02f34a2634aa4de265cfe4af5f0","ref":"refs/heads/main","pushedAt":"2024-05-28T21:40:19.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"use total_L is None for dense_to_jagged + fix linter errors (#2599)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2599\n\n# Could not guard on data-dependent expression Ne(u0, 0)\n\n```\nRuntimeError: Failed running call_function fbgemm.dense_to_jagged(*(FakeTensor(..., device='cuda:0', size=(s1, 2540, 512), dtype=torch.bfloat16), [FakeTensor(..., device='cuda:0', size=(s1 + 1,), dtype=torch.int64)]), **{'total_L': None}):\nCould not guard on data-dependent expression Ne(u0, 0) (unhinted: Ne(u0, 0)). (Size-like symbols: u0)\nATTENTION: guard_size_oblivious would fix the error, evaluating expression to True.\nMaybe you need to add guard_size_oblivious to framework code, see doc below for more guidance.\nPotential framework code culprit (scroll up for full backtrace):\n File \"/mnt/xarfuse/uid-119376/a7f0c177-seed-nspid4026531836_cgpid3083025-ns-4026531841/fbgemm_gpu/sparse_ops.py\", line 467, in dense_to_jagged_forward\n if not total_L:\n\nUser Stack (most recent call last):\n (snipped, see stack below for prefix)\n File \".82\", line 424, in forward\n dense_to_jagged = torch.ops.fbgemm.dense_to_jagged(matmul_22, [asynchronous_complete_cumsum_9], total_L = None); matmul_22 = None\n```\n\nReviewed By: frank-wei\n\nDifferential Revision: D57465638\n\nfbshipit-source-id: 2cf9acbedfeeef0548d8f963a85a91e53424c71c","shortMessageHtmlLink":"use total_L is None for dense_to_jagged + fix linter errors (#2599)"}},{"before":"668eba3fb98b88a911231db71eb24804807a9570","after":"5bd335fd0f8f0fce831c7fb49f96d970ec95c96d","ref":"refs/heads/main","pushedAt":"2024-05-28T20:27:07.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Add AMD CK FP8 Kernels to Llama Dispatch (#2630)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2630\n\nThis diff replaces the slow `torch._scaled_mm` implementation with faster CK kernels added in D56963018. I also extend AMD support to rowwise fp8 quantization.\n\nPreviously we didn't have an accelerated kernel to quantize tensors to fp8 on AMD, I updated the Triton functions to support AMD and use them for rowwise quantization and get excellent performance.\n\nBefore this change we have:\n```\nINFO:root:BF16 T: 5966.47us, FLOPS: 483.74TF/s\nINFO:root:BF16 (G) T: 6017.79us, FLOPS: 479.61TF/s\nINFO:root:FP8D AMD T: 32761.27us, FLOPS: 88.10TF/s\nINFO:root:FP8D AMD (G) T: 33014.49us, FLOPS: 87.42TF/s\n```\n\nAfter we have:\n```\nINFO:root:BF16 T: 6006.43us, FLOPS: 480.52TF/s\nINFO:root:BF16 (G) T: 6045.48us, FLOPS: 477.42TF/s\nINFO:root:FP8D AMD CK T: 5894.74us, FLOPS: 489.63TF/s\nINFO:root:FP8D AMD CK (G) T: 5870.20us, FLOPS: 491.67TF/s\nINFO:root:FP8D rowwise AMD CK T: 3877.73us, FLOPS: 744.31TF/s\nINFO:root:FP8D rowwise AMD CK (G) T: 3892.07us, FLOPS: 741.56TF/s\n```\n\nWhen using LLama3 shapes the performance is:\n```\nINFO:root:BF16 T: 90416.92us, FLOPS: 474.26TF/s\nINFO:root:BF16 (G) T: 91022.73us, FLOPS: 471.10TF/s\nINFO:root:FP8D AMD CK T: 65453.99us, FLOPS: 655.13TF/s\nINFO:root:FP8D AMD CK (G) T: 63632.00us, FLOPS: 673.89TF/s\nINFO:root:FP8D rowwise AMD CK T: 60791.51us, FLOPS: 705.38TF/s\nINFO:root:FP8D rowwise AMD CK (G) T: 60916.44us, FLOPS: 703.93TF/s\n```\n\nReviewed By: choutim\n\nDifferential Revision: D57739339\n\nfbshipit-source-id: fb2c38e25c0c349810d5e968218911508c1ccd97","shortMessageHtmlLink":"Add AMD CK FP8 Kernels to Llama Dispatch (#2630)"}},{"before":"9f1d0ef86d26f61e708b4e0ad851b004e347158e","after":"668eba3fb98b88a911231db71eb24804807a9570","ref":"refs/heads/main","pushedAt":"2024-05-28T19:26:03.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"SymInt-ify Optimizer NONE: total_unique_indices arg (#2635)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2635\n\n1/ To be able to PT2 compile OptimType.NONE - we need to make total_unique_indices SymInt.\n\nIt's generated via python generator - introducing SymInt arg type\n\n2/ adding TORCH_GUARD_SIZE_OBLIVIOUS to pass through data dependent conditions\n\nReviewed By: q10\n\nDifferential Revision: D57861244\n\nfbshipit-source-id: 471b4f353fc5e6c52759f66b7b76940743cf1f84","shortMessageHtmlLink":"SymInt-ify Optimizer NONE: total_unique_indices arg (#2635)"}},{"before":"9df527cbc974b2e8921cc15cfe249c834807a2fc","after":"6d797095e2054eb35f688df54b9ed4e1348f2c0a","ref":"refs/heads/nightly","pushedAt":"2024-05-28T11:34:45.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-28 nightly release (9f1d0ef86d26f61e708b4e0ad851b004e347158e)","shortMessageHtmlLink":"2024-05-28 nightly release (9f1d0ef)"}},{"before":"af8537fd20ced2dc90df57b7d1a03e713e47ef35","after":"79ff8396fb5c9995ef271db520675d2aa5b44866","ref":"refs/heads/gh-pages","pushedAt":"2024-05-28T01:51:22.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@9f1d0ef86d26f61e708b4e0ad851b004e347158e ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 9f1d0ef ๐Ÿš€"}},{"before":"5dd515eeb3c22e49e426c5487ce04382285c9544","after":"9f1d0ef86d26f61e708b4e0ad851b004e347158e","ref":"refs/heads/main","pushedAt":"2024-05-28T01:39:42.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Add TBE tests to failures_dict_fast.json (#2632)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2632\n\nAs title\n\nReviewed By: q10, spcyppt\n\nDifferential Revision: D57758099\n\nfbshipit-source-id: d90aaa5ce6461251e53f08074504b4cc2738eaab","shortMessageHtmlLink":"Add TBE tests to failures_dict_fast.json (#2632)"}},{"before":"b6ddc32a699a357b02ecc5b4bf77e94b015a6e84","after":"9df527cbc974b2e8921cc15cfe249c834807a2fc","ref":"refs/heads/nightly","pushedAt":"2024-05-27T11:35:34.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-27 nightly release (5dd515eeb3c22e49e426c5487ce04382285c9544)","shortMessageHtmlLink":"2024-05-27 nightly release (5dd515e)"}},{"before":"5b01c5300d1417f4a266023da1f20486b3b2b723","after":"b6ddc32a699a357b02ecc5b4bf77e94b015a6e84","ref":"refs/heads/nightly","pushedAt":"2024-05-26T11:35:06.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-26 nightly release (5dd515eeb3c22e49e426c5487ce04382285c9544)","shortMessageHtmlLink":"2024-05-26 nightly release (5dd515e)"}},{"before":"b8961051d323b5b2e18d29c74c5756af4883b27f","after":"5b01c5300d1417f4a266023da1f20486b3b2b723","ref":"refs/heads/nightly","pushedAt":"2024-05-25T11:34:44.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-25 nightly release (5dd515eeb3c22e49e426c5487ce04382285c9544)","shortMessageHtmlLink":"2024-05-25 nightly release (5dd515e)"}},{"before":"acca87ba07a33607c51728a3993d01b6cb802187","after":"af8537fd20ced2dc90df57b7d1a03e713e47ef35","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T23:52:24.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@5dd515eeb3c22e49e426c5487ce04382285c9544 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 5dd515e ๐Ÿš€"}},{"before":"db4d37961c9943e37569f71388d52eac2682d83f","after":"5dd515eeb3c22e49e426c5487ce04382285c9544","ref":"refs/heads/main","pushedAt":"2024-05-24T23:42:40.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Use do_bench_cudagraph for stable benchmarking (#2633)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2633\n\n`do_bench_cudagraph` is known more stable than `do_bench`. Let's use it for benchmarking.\n\nReviewed By: sryap\n\nDifferential Revision: D57799962\n\nfbshipit-source-id: c98879d6a42b3ad94e7f39cd1bdf8362a62cebc8","shortMessageHtmlLink":"Use do_bench_cudagraph for stable benchmarking (#2633)"}},{"before":"06fe4ad68b3aa979893eca1b6b3696dcc3b2ec7f","after":"acca87ba07a33607c51728a3993d01b6cb802187","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T22:01:39.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@db4d37961c9943e37569f71388d52eac2682d83f ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ db4d379 ๐Ÿš€"}},{"before":"ab05ca9846186da9d705a43fb1a55fe3cdbc27a9","after":"db4d37961c9943e37569f71388d52eac2682d83f","ref":"refs/heads/main","pushedAt":"2024-05-24T21:52:07.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Add cache conflict miss support (#2596)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2596\n\nPrior to this diff, SSD TBE lacked support for the conflict cache miss\nscenario. It operated under the assumption that the cache, located in\nGPU memory, was sufficiently large to hold all prefetched data from\nSSD. In the event of a conflict cache miss, the behavior of SSD TBE\nwould be unpredictable (it could either fail or potentially access\nillegal memory). Note that a conflict cache miss happens when an\nembedding row is absent in the cache, and after being fetched from\nSSD, it cannot be inserted into the cache due to capacity constraints\nor associativity limitations.\n\nThis diff introduces support for conflict cache misses by storing rows\nthat cannot be inserted into the cache due to conflicts in a scratch\npad, which is a temporary GPU tensor. In the case where rows are\nmissed from the cache, TBE kernels can access the scratch pad.\n\nPrior to this diff, during the SSD prefetch stage, any row that was\nmissed the cache and required fetching from SSD would be first fetched\ninto a CPU scratch pad and then transferred to GPU. Rows that could be\ninserted into the cache would subsequently be copied from the GPU\nscratch pad into the cache. If conflict misses occurred, the prefetch\nbehavior would be unpredictable. With this diff, conflict missed rows\nare now retained in the scratch pad, which is kept alive until the\ncurrent iteration completes. Throughout the forward and backward +\noptimizer stages of TBE, both the cache and scratch pad are equivalent\nin terms of usage. However, following the completion of the backward +\noptimizer step, rows in the scratch pad are flushed back to SSD,\nunlike rows residing in the cache which are not evicted for future\nusage (see the diagram below for more details).\n\n {F1645878181}\n\nReviewed By: spcyppt\n\nDifferential Revision: D55998215\n\nfbshipit-source-id: 79846dd7ae0ec95752fa1cb427790f685900bf6e","shortMessageHtmlLink":"Add cache conflict miss support (#2596)"}},{"before":"57978a07b573d16df843f37b628f980fca3babf2","after":"06fe4ad68b3aa979893eca1b6b3696dcc3b2ec7f","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T17:48:52.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@ab05ca9846186da9d705a43fb1a55fe3cdbc27a9 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ ab05ca9 ๐Ÿš€"}},{"before":"63ca6dcedc501f66cd1306c1b57817d9842239f0","after":"ab05ca9846186da9d705a43fb1a55fe3cdbc27a9","ref":"refs/heads/main","pushedAt":"2024-05-24T17:39:14.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"permute_2D_sparse_data Autograd formula (#2629)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2629\n\nReland of D57625720 without dependency on torchrec\n\nAdding permute_2D_sparse_data python formula for Inductor compilation.\n\nReviewed By: ezyang\n\nDifferential Revision: D57773001\n\nfbshipit-source-id: bf73e5b79c0450cdcc123eb4630941d668beb1f9","shortMessageHtmlLink":"permute_2D_sparse_data Autograd formula (#2629)"}},{"before":"b417156b313189b595e488fff865733557211e32","after":"b8961051d323b5b2e18d29c74c5756af4883b27f","ref":"refs/heads/nightly","pushedAt":"2024-05-24T11:35:17.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"pytorchbot","name":null,"path":"/pytorchbot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/21957446?s=80&v=4"},"commit":{"message":"2024-05-24 nightly release (63ca6dcedc501f66cd1306c1b57817d9842239f0)","shortMessageHtmlLink":"2024-05-24 nightly release (63ca6dc)"}},{"before":"65648f71d60b58ea14a144fac3e5281c8e7d9d61","after":"57978a07b573d16df843f37b628f980fca3babf2","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T02:33:02.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@63ca6dcedc501f66cd1306c1b57817d9842239f0 ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 63ca6dc ๐Ÿš€"}},{"before":"42850883b2793ab0ffa8d47296182c259658095c","after":"63ca6dcedc501f66cd1306c1b57817d9842239f0","ref":"refs/heads/main","pushedAt":"2024-05-24T02:23:29.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Add INT4-FP8 rowwise matmul tests (#2622)\n\nSummary:\nPull Request resolved: https://github.com/pytorch/FBGEMM/pull/2622\n\n- Add INT4-FP8 rowwise matmul tests with cudagraph and eager\n- Change number of groups to group size to align with E2E variable\n- Fix a minor dtype issue\n\nReviewed By: jwfromm\n\nDifferential Revision: D57677844\n\nfbshipit-source-id: 78578490025d72a5d73dd207fce4518304283de0","shortMessageHtmlLink":"Add INT4-FP8 rowwise matmul tests (#2622)"}},{"before":"f3620f012e9665c0c63f4e66452782d1437a24c7","after":"65648f71d60b58ea14a144fac3e5281c8e7d9d61","ref":"refs/heads/gh-pages","pushedAt":"2024-05-24T02:16:13.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@42850883b2793ab0ffa8d47296182c259658095c ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 4285088 ๐Ÿš€"}},{"before":"0ac664b0e2b1128816be1b78f9f66346f6dec36f","after":"42850883b2793ab0ffa8d47296182c259658095c","ref":"refs/heads/main","pushedAt":"2024-05-24T02:06:32.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"facebook-github-bot","name":"Facebook Community Bot","path":"/facebook-github-bot","primaryAvatarUrl":"https://avatars.githubusercontent.com/u/6422482?s=80&v=4"},"commit":{"message":"Revert D57625720: permute_2D_sparse_data Autogad formula\n\nDifferential Revision:\nD57625720\n\nOriginal commit changeset: 4596111afb12\n\nOriginal Phabricator Diff: D57625720\n\nfbshipit-source-id: 99725dd01245c157b6cec0212c7d2a0a10718bb2","shortMessageHtmlLink":"Revert D57625720: permute_2D_sparse_data Autogad formula"}},{"before":"ee33cb5a4972c846a9b8fb5c934b99f9ee94b00f","after":"f3620f012e9665c0c63f4e66452782d1437a24c7","ref":"refs/heads/gh-pages","pushedAt":"2024-05-23T23:50:21.000Z","pushType":"push","commitsCount":1,"pusher":{"login":"github-actions[bot]","name":null,"path":"/apps/github-actions","primaryAvatarUrl":"https://avatars.githubusercontent.com/in/15368?s=80&v=4"},"commit":{"message":"Deploying to gh-pages from @ pytorch/FBGEMM@0ac664b0e2b1128816be1b78f9f66346f6dec36f ๐Ÿš€","shortMessageHtmlLink":"Deploying to gh-pages from @ 0ac664b ๐Ÿš€"}}],"hasNextPage":true,"hasPreviousPage":false,"activityType":"all","actor":null,"timePeriod":"all","sort":"DESC","perPage":30,"cursor":"djE6ks8AAAAEV6F5PwA","startCursor":null,"endCursor":null}},"title":"Activity ยท pytorch/FBGEMM"}