Skip to content

feature: add support for native PT DDP distribution #236

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 1 commit into from
Jul 29, 2022
Merged

feature: add support for native PT DDP distribution #236

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 10 additions & 1 deletion src/sagemaker_pytorch_container/training.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -21,6 +21,8 @@

MASTER_PORT = '7777'
LAUNCH_SMDATAPARALLEL_ENV_NAME = 'sagemaker_distributed_dataparallel_enabled'
LAUNCH_MPI_ENV_NAME = 'sagemaker_mpi_enabled'
LAUNCH_PYTORCH_DDP_ENV_NAME = "sagemaker_pytorch_ddp_enabled"

logger = logging.getLogger(__name__)

Expand Down Expand Up @@ -49,7 +51,11 @@ def train(training_environment):

_set_distributed_environment(training_environment)

mpi_enabled = training_environment.additional_framework_parameters.get('sagemaker_mpi_enabled')
mpi_enabled = training_environment.additional_framework_parameters.get(LAUNCH_MPI_ENV_NAME)

pytorch_ddp_enabled = training_environment.additional_framework_parameters.get(
LAUNCH_PYTORCH_DDP_ENV_NAME, False
)

smdataparallel_enabled = training_environment.additional_framework_parameters.get(
LAUNCH_SMDATAPARALLEL_ENV_NAME, False
Expand All @@ -60,6 +66,9 @@ def train(training_environment):
if training_environment.current_instance_group in training_environment.distribution_instance_groups:
if mpi_enabled:
runner_type = runner.MPIRunnerType
elif pytorch_ddp_enabled:
runner_type = runner.SMDataParallelRunnerType
logger.info('Invoking SMDataParallel for native PT DDP job')
elif smdataparallel_enabled:
runner_type = runner.SMDataParallelRunnerType
logger.info('Invoking SMDataParallel')
Expand Down
16 changes: 16 additions & 0 deletions test/unit/test_train.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -90,6 +90,22 @@ def test_train_smdataparallel(run_module, training_env):
)


@patch("sagemaker_training.entry_point.run")
@patch('socket.gethostbyname', MagicMock())
def test_train_pytorch_ddp(run_module, training_env):
training_env.additional_framework_parameters["sagemaker_pytorch_ddp_enabled"] = True

train(training_env)
run_module.assert_called_with(
uri=training_env.module_dir,
user_entry_point=training_env.user_entry_point,
args=training_env.to_cmd_args(),
env_vars=training_env.to_env_vars(),
capture_error=True,
runner_type=runner.SMDataParallelRunnerType,
)


@patch('sagemaker_training.entry_point.run', MagicMock())
@patch('socket.gethostbyname', MagicMock())
def test_environment(training_env):
Expand Down