sanshang 249152475d fix sequence number for group (#134578) ... Summary: Fix sequence number in execution trace dump for matching between collective/p2p op and wait in execution trace replay. `ProcessGroupNCCL` has 2 sequence number counter, `seqCollective_` and `seqP2P_`. b18ba9419e/torch/csrc/distributed/c10d/ProcessGroupNCCL.hpp (L1188-L1191) However, `WorkNCCL` only has one sequence number member `seq_`. b18ba9419e/torch/csrc/distributed/c10d/ProcessGroupNCCL.hpp (L387) We need to match collective and p2p with wait separately. 29b5a462dc Depend on: https://github.com/pytorch/pytorch/pull/135132 Test Plan: buck2 run mode/dev-nosan kineto/libkineto/fb/integration_tests:pytorch_execution_trace_integration_test Differential Revision: Pull Request resolved: https://github.com/pytorch/pytorch/pull/134578 Approved by: https://github.com/kwen2501, https://github.com/c-p-i-o		2024-10-10 04:24:06 +00:00
..
aoti_abi_check
aoti_inference
api	[Environment Variable][2/N] Use thread-safe setenv wrapper (#124485)	2024-10-04 07:30:51 +00:00
c10d	fix sequence number for group (#134578)	2024-10-10 04:24:06 +00:00
common
dist_autograd
jit	Revert "Fix clang-tidy warnings in Caffe2 code (#134935)"	2024-09-13 16:42:37 +00:00
lazy
lite_interpreter_runtime
monitor
profiler
rpc
tensorexpr
__init__.py