8000 Commits · n0k0m3/pytorch · GitHub

More Web Proxy on the site http://driver.im/

Commits

Commits on Mar 19, 2025

[c10d] Add a collective time estimator for NCCL comms (#149343)

fduwjj
authored and
pytorchmergebot
committed

Commits on Mar 14, 2025

[c10d] Add param recording for uniqueID broadcasting and allgather (#149166)

fduwjj
authored and
pytorchmergebot
committed

Commits on Mar 5, 2025

[c10d] Move record param for init to the right place (#148571)

fduwjj
authored and
pytorchmergebot
committed

Commits on Feb 19, 2025

[fr][fix] Split MatchState and dynamic info for fr analysis downstream (#147439)

fduwjj
authored and
pytorchmergebot
committed

Commits on Jan 31, 2025

[c10d][NCCL] Implement ncclCommInitRankScalable (merging #136789) (#144794)
10000

fduwjj
authored and
pytorchmergebot
committed

Commits on Jan 28, 2025

[c10d][ez] Remove goto in PGNCCL and make linter happy for PGNCCL and NCCLUtils (#145855)

fduwjj
authored and
pytorchmergebot
committed

Commits on Jan 16, 2025

[c10d][fr] Fix the bug when we still mark mismatch when there are match case (#144916)

fduwjj
authored and
pytorchmergebot
committed

Commits on Jan 15, 2025

[c10d] Fix CudaEventCache for dangling references (#144496)

fduwjj
authored and
pytorchmergebot
committed
[c10d][ez] Add comments to the end of Macro for better readability (#144789)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 27, 2024

[FR] Include mismatch rank into mismatch_collectives and update log message (#141631)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 26, 2024

[c10d] Enable CudaEventCache by default and add multi device support (#140975)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 21, 2024

[c10d] Switch all timer logging in c10d to wait_counter (#141154)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 13, 2024

Commits on Nov 12, 2024

[FR] refactor build collective and return more info to db (#140082) (#140303)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 11, 2024

[FR] Enable best effort parital analysis and verbose mode for trace printing (#139853)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 9, 2024

[c10d][Logging] Remove args and kwargs from c10d logging (#140169)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 8, 2024

[c10d][Logging] Remove args and kwargs from c10d logging (#140169)

fduwjj
authored and
pytorchmergebot
committed
Fix for T206766523 ("Your diff, D65462767, broke some tests") (#139804)

fduwjj
authored and
pytorchmergebot
committed

Commits on Nov 5, 2024

Commits on Oct 24, 2024

[c10d] Reorder GIL checker and c++ stack trace print with comments (#138734)

fduwjj
authored and
pytorchmergebot
committed

Commits on Oct 19, 2024

[c10d] Reorder cpp stack dump and FR dump and add log prefix to loggings (#138368)

fduwjj
authored and
pytorchmergebot
committed

Commits on Oct 16, 2024

Commits on Oct 3, 2024

[FR] Fix duplicate output for the case when not all ranks join on collective (#137256)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 29, 2024

[Elasitc][fix] Use the right env variable TORCH_ELASTIC_WORKER_IDENTICAL for unit test (#136916)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 27, 2024

[reland][Elastic] Skip store barrier and store get in host assign (#136865)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 26, 2024

[reland] [torchelastic][c10d] Fix store prefix race in rendezvous (#136768)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 24, 2024

[FR] Make OSS fr_trace function available for internal script and improve pg filtering (#136473)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 23, 2024

[RFC][torchelastic][c10d] Fix store prefix race in rendezvous (#135957)

fduwjj
authored and
pytorchmergebot
committed
[FR] Fix errors in all2all check, improve some log output (#136399)

fduwjj
authored and
pytorchmergebot
committed

Commits on Sep 18, 2024

[c10d] Make test compatible for new pytest (#136158)

fduwjj
authored and
pytorchmergebot
committed

0