Ifu 2023 12 14 #52
Merged
Ifu 2023 12 14 #52
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Summary: As titled Pull Request resolved: pytorch#2192 Reviewed By: q10 Differential Revision: D51912522 fbshipit-source-id: 1eb3bad0bf0b1cb3ee45ea74572115dcd971fb0a
Summary: D51921135 added PT2 support for mutable operators. This added support for the following operators: - fbgemm::bounds_check_indices - fbgemm::direct_mapped_lru_cache_populate_byte - fbgemm::emb_inplace_update - fbgemm::lru_cache_populate_byte - fbgemm::lxu_cache_flush - fbgemm::lxu_cache_locking_counter_decrement - fbgemm::pruned_hashmap_insert so we need to update the test failures (those tests are no longer failing). Reviewed By: bdhirsh Differential Revision: D51942915 fbshipit-source-id: 867e791b7e86c0a72ea7d88837a56eb8b7f6cc48
Summary: Pull Request resolved: pytorch#2190 Add autogenerated opcheck tests to `permute_pooled_embedding_test.py`. Add a test for `fbgemm::permute_pooled_embs`. Add PT2 compliant tag to `fbgemm::permute_pooled_embs[_auto_grad]` since they pass the opcheck tests. Reviewed By: zou3519 Differential Revision: D51875123 fbshipit-source-id: 497ee9a1ac357c302e3394b4053c238a06d5a0d0
Summary: Pull Request resolved: pytorch#2193 1. Remove the meta cpp function fbgemm::dense_to_jagged() and fbgemm::dense_to_jagged_forward() 2. Replace it with the Python abstract impl dense_to_jagged() Reviewed By: zou3519, yanboliang Differential Revision: D51216256 fbshipit-source-id: 532f25e5f9574e75f310ce98325523eb684cc7c8
Summary: - Add support for Python 3.12 Blocked until pytorch/pytorch#110436 is resolved See [reference](https://github.com/pytorch/vision/pull/8137/files) Pull Request resolved: pytorch#2194 Reviewed By: spcyppt Differential Revision: D51929976 Pulled By: q10 fbshipit-source-id: 4f04fe9af6ea8f37ef4fbe7078618bd104826ac8
Summary: Pull Request resolved: pytorch#2199 Forward fix on removing expected failures of all dense_to_jagged tests as they have all been fixed by D51216256 Reviewed By: jspark1105, zou3519 Differential Revision: D51959191 fbshipit-source-id: f71e8cc43a738267cef9b15a768f38ccbf2839c5
Summary: Pull Request resolved: pytorch#2195 Instead of using the ref implementation for sequence embedding on CPU, this diff directs TBE to invoke the AVX implementation of pooled TBE by forcing pooling factors of 1 (i.e., passing `at::arange(index_size + 1)` as offfsets). The performance gained from using the AVX implementation offsets the overhead incurs in creating the new offsets. Reviewed By: jspark1105, YazhiGao Differential Revision: D51918878 fbshipit-source-id: cc853534cb42ff0c2083ba37d87d2daed12b1efb
…h#2196) Summary: Pull Request resolved: pytorch#2196 Consolidate the function calls using Jinja macro Reviewed By: jspark1105 Differential Revision: D51922356 fbshipit-source-id: 8aea70a43f09a980c12eb89932d8d25bac2d7ef2
Summary: Pull Request resolved: pytorch#2200 As title Reviewed By: YazhiGao Differential Revision: D51963691 fbshipit-source-id: 45604dc2a7e4a029bc6172da4ef39d7ab648dc34
Summary: Pull Request resolved: pytorch#2189 Reviewed By: q10 Differential Revision: D51609251 fbshipit-source-id: d6ae1a74ebfa9b207b271866a157f73a0b562298
Summary: Pull Request resolved: pytorch#2201 These pass all of the generated tests. Reviewed By: williamwen42 Differential Revision: D51993091 fbshipit-source-id: c7083a6c8a4735100f55334653f3a21e65dfaa83
…liant (pytorch#2198) Summary: Pull Request resolved: pytorch#2198 In particular `split_embedding_codegen_lookup_rowwise_adagrad_function`. But a bunch of similar ops can be marked as pt2_compliant by fixing some templates and bugs. Reviewed By: zou3519 Differential Revision: D51960321 fbshipit-source-id: 0d15eedd0aa4fb78d8d7ecc88fc170b648a26a13
Summary: - Add script to collect PyTorch environment information for use when reporting issues to the PyTorch team Pull Request resolved: pytorch#2203 Reviewed By: spcyppt Differential Revision: D52046632 Pulled By: q10 fbshipit-source-id: 1c3052bc00ab9ebaccb547568ce45b111b5082a2
Summary: OSS CI jobs often failed at checkout step with error `Error: File was unable to be removed Error: EACCES: permission denied, rmdir '/home/ec2-user/actions-runner/_work/FBGEMM/FBGEMM/3'` This is because during the build process of one job, the script executes installations of dependencies in the folder under different user (ie.e,`root` in Nova job). In the subsequent job run on the same runner, `ec2-user` tries to clean up the folder but unable able to, causing the permission denied error. For example, https://github.com/pytorch/FBGEMM/actions/runs/7151275180/job/19475677899 all failed because the runners were used to build wheel in the previous job, i.e. https://github.com/pytorch/FBGEMM/actions/runs/7150909712/job/19474677106. Pull Request resolved: pytorch#2204 Reviewed By: q10, huydhn Differential Revision: D52054787 fbshipit-source-id: f363e1c948ffe9b95f1d384bbca521faa78c12b9
Summary: Pull Request resolved: pytorch#2202 This diff adds an early exit in sparse_async_cumsum ops. When the size(s) of input tensor are zeor(s) ops return zero tensor. Reviewed By: jspark1105, sryap, jasonjk-park Differential Revision: D51999938 fbshipit-source-id: 5c2151a01f547666a1e1742f334953c01dbf6630
Summary: - Add debug flags for HIP runs when testing - Fix FBGEMM_GPU-ROCm building process for ROCm 5.7 - Redo and reduce the workflow definitions for ROCm runs Pull Request resolved: pytorch#2206 Reviewed By: sryap Differential Revision: D52099327 Pulled By: q10 fbshipit-source-id: 98c6fae53ebfdc54ae9fa6cc9997166651e03454
Summary: - Re-organize the OSS workflows to de-duplicate jobs Pull Request resolved: pytorch#2209 Reviewed By: spcyppt Differential Revision: D52102704 Pulled By: q10 fbshipit-source-id: 6eca3b0c05cdcfb80d06e9e2fad0a96b851d8879
… exit in sparse_async_cumsum ops" for otest failure (pytorch#2208) Summary: Pull Request resolved: pytorch#2208 This diff is reverting D51999938 D51999938: Add early exit in sparse_async_cumsum ops by meremeev has been identified to be causing the following test failure: Tests affected: - [deeplearning/fbgemm/fbgemm_gpu:sparse_ops_test - test_schema__test_asynchronous_complete_cumsum_2d (deeplearning.fbgemm.fbgemm_gpu.test.sparse_ops_test.SparseOpsTest)](https://www.internalfb.com/intern/test/562950068047718/) Here's the Multisect link: https://www.internalfb.com/multisect/3757962 Here are the tasks that are relevant to this breakage: We're generating a revert to back out the changes in this diff, please note the backout may land if someone accepts it. If you believe this diff has been generated in error you may Commandeer and Abandon it. Reviewed By: jasonjk-park Differential Revision: D52099677 fbshipit-source-id: 4e36745864148e5bb337465b0a9afcfe80846389
Summary: Pull Request resolved: pytorch#2207 `-Wextra-semi` or `-Wextra-semi-stmt` If the code compiles, this is safe to land. Reviewed By: palmje Differential Revision: D51995043 fbshipit-source-id: 4a263444eb5c2086c954de09c8a6c85ca507e34c
Summary: - Add compile flags to enable device-side assertions Pull Request resolved: pytorch#2211 Reviewed By: spcyppt Differential Revision: D52143964 Pulled By: q10 fbshipit-source-id: 7c4b7bbe0f85e390a4fc334abbf0361ca0c9db42
Summary: Pull Request resolved: pytorch#2212 The momentum1_host is a write back tensor and in the functiona schema, it should be labeled as "Tensor(b!)" in order to make the CPU fallback correctly write data back to it. Can't find a more elegant solution to fix this now. Reviewed By: jspark1105 Differential Revision: D52082756 fbshipit-source-id: f19ad2332ec5a0f150ad37e7203cb8682fad26a6
Summary: As titled. Pull Request resolved: pytorch#2210 Reviewed By: q10 Differential Revision: D52107179 fbshipit-source-id: c5d2e9de6edd8a072d03d731474ac574270ba78d
…into IFU-2023-12-14
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.