Tags: pruthvistony/pytorch
Tags
Update pthreadpool to pthreadpool:029c88620802e1361ccf41d1970bd5b07fd… …6b7bb. (pytorch#40524) (pytorch#41190) Summary: Pull Request resolved: pytorch#40524 Reviewed By: ezyang Differential Revision: D22215742 Pulled By: AshkanAliabadi fbshipit-source-id: ef594e0901337a92b21ddd44e554da66c723eb7c
Release GIL during DDP construction. (pytorch#40877) Summary: Pull Request resolved: pytorch#40495 As part of debugging flaky ddp_under_dist_autograd tests, I realized we were running into the following deadlock. 1) Rank 0 would go into DDP construction, hold GIL and wait for broadcast in DDP construction. 2) Rank 3 is a little slower and performs an RRef fetch call before the DDP construction. 3) The RRef fetch call is done on Rank 0 and tries to acquire GIL. 4) We now have a deadlock since Rank 0 is waiting for Rank 3 to enter the collective and Rank 3 is waiting for Rank 0 to release GIL. ghstack-source-id: 106534442 Test Plan: 1) Ran ddp_under_dist_autograd 500 times. 2) waitforbuildbot Differential Revision: D22205180 fbshipit-source-id: 6afd55342e801b9edb9591ff25158a244a8ea66a Co-authored-by: Pritam Damania <pritam.damania@fb.com>
.circleci: Fix upload to backup directory Signed-off-by: Eli Uriegas <eliuriegas@fb.com>
[ONNX] Fix pow op export [1.5.1] (pytorch#39791) * [ONNX] Fix pow op export (pytorch#38065) Summary: Fix pow type cast for opset 9 and update opset 12 Pull Request resolved: pytorch#38065 Differential Revision: D21485353 Pulled By: malfet fbshipit-source-id: 3993e835ffad07b2e6585eb5cf1cb7c8474de2ec * Update ort-nighly version as suggested in pytorch#39685 (comment) * Apply changes from pytorch#37846 to `test_topk_smallest_unsorted` Co-authored-by: neginraoof <neginmr@utexas.edu>
[ONNX] Fix pow op export [1.5.1] (pytorch#39791) * [ONNX] Fix pow op export (pytorch#38065) Summary: Fix pow type cast for opset 9 and update opset 12 Pull Request resolved: pytorch#38065 Differential Revision: D21485353 Pulled By: malfet fbshipit-source-id: 3993e835ffad07b2e6585eb5cf1cb7c8474de2ec * Update ort-nighly version as suggested in pytorch#39685 (comment) * Apply changes from pytorch#37846 to `test_topk_smallest_unsorted` Co-authored-by: neginraoof <neginmr@utexas.edu>
[v.1.5.0] Ensure linearIndex of advanced indexing backwards is contig… ( pytorch#36962) * [v.1.5.0] Ensure linearIndex of advanced indexing backwards is contiguous. This is a more straightforward solution to the problem than pytorch#36957; I don't know about the relative performance. Fixes: pytorch#36956 ghstack-source-id: 43c48ea Pull Request resolved: pytorch#36959 * Fix test.
[v.1.5.0] Ensure linearIndex of advanced indexing backwards is contig… ( pytorch#36962) * [v.1.5.0] Ensure linearIndex of advanced indexing backwards is contiguous. This is a more straightforward solution to the problem than pytorch#36957; I don't know about the relative performance. Fixes: pytorch#36956 ghstack-source-id: 43c48ea Pull Request resolved: pytorch#36959 * Fix test.
Use counter instead of vector of futures in `_parallel_run` (pytorch#… …36159) (pytorch#36334) Summary: This should be faster than allocating one mutex, flag and conditional variable per task. Using `std::atomic<size_t>` to count remaing tasks is not sufficient, because modification of remaining counter and signalling conditional variable must happen atomically, otherwise `wait()` might get invoked after `notify_one()` was called. Pull Request resolved: pytorch#36159 Test Plan: CI Differential Revision: D20905411 Pulled By: malfet fbshipit-source-id: facaf599693649c3f43edafc49f369e90d2f60de (cherry picked from commit 986a8fd) Signed-off-by: Eli Uriegas <eliuriegas@fb.com> Co-authored-by: Nikita Shulga <nshulga@fb.com>
Revert "Fix handling of non-finite values in topk (pytorch#35253)" (p… …ytorch#35582) This reverts commit b12579d. This patch in-and-of itself looks fine, but it's causing some AMP tests to fail.
PreviousNext