Update distributed contributing guide to show how to run one test in test_distributed_spawn (#67801)

Summary:
Running one test in test_distributed_spawn is a bit confusing but possible. Add documentation to the CONTRIBUTING.md for this.

cc pietern mrshenli pritamdamania87 zhaojuanmao satgera rohan-varma gqchen aazzolini osalpekar jiayisuse SciPioneer H-Huang

Pull Request resolved: https://github.com/pytorch/pytorch/pull/67801

Reviewed By: mrshenli

Differential Revision: D32157700

Pulled By: rohan-varma

fbshipit-source-id: a1d10f2fb5f169b46c6d15149bf949082d9bd200
This commit is contained in:
Rohan Varma 2021-11-04 08:53:06 -07:00 committed by Facebook GitHub Bot
parent 4262c8913c
commit bd8feb33d4

View File

@ -87,6 +87,9 @@ python test/distributed/test_pg_wrapper.py
# Run distributed tests, including tests for Distributed Data Parallel.
python test/run_test.py --verbose -i distributed/test_distributed_spawn
# Run a single test in the test_distributed_spawn test suite.
touch /tmp/barrier && TEMP_DIR="/tmp" BACKEND="nccl" WORLD_SIZE="2" python test/distributed/test_distributed_spawn.py -v TestDistBackendWithSpawn.test_ddp_profiling_torch_profiler
# Run the RPC test suite for the TensorPipeAgent.
python test/distributed/rpc/test_tensorpipe_agent.py
python test/distributed/rpc/cuda/test_tensorpipe_agent.py