mirror of
https://github.com/saymrwulf/pytorch.git
synced 2026-05-14 20:57:59 +00:00
The resources directory lets ET observer dump any additional data like Triton kernels while capturing the ET. This allows us to use the ET trace to replay PT2 workloads and get visibility into data like generated kernels and their usage in a model, index tensor data etc. We also added a few ways to enable ET and ET Resources through the OS environment variables. Setting `ENABLE_PYTORCH_EXECUTION_TRACE` will enable default Execution Tracing in Pytorch. Additionally setting `ENABLE_PYTORCH_EXECUTION_TRACE_EXTRAS` will enable ET to collect extra resources from the ET run like Triton Kernels. Differential Revision: [D67610542](https://our.internmc.facebook.com/intern/diff/D67610542/) **NOTE FOR REVIEWERS**: This PR has internal Meta-specific changes or comments, please review them on [Phabricator](https://our.internmc.facebook.com/intern/diff/D67610542/)! Pull Request resolved: https://github.com/pytorch/pytorch/pull/143775 Approved by: https://github.com/shengfukevin, https://github.com/wdvr |
||
|---|---|---|
| .. | ||
| profiler_utils_mock_events.json | ||
| test_cpp_thread.cpp | ||
| test_cpp_thread.py | ||
| test_cpp_thread_lib.pyi | ||
| test_execution_trace.py | ||
| test_kineto.py | ||
| test_memory_profiler.py | ||
| test_profiler.py | ||
| test_profiler_tree.py | ||
| test_record_function.py | ||
| test_torch_tidy.py | ||