Skip to content
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

Fix unfixed naming inconsistencies #531

Merged
merged 1 commit into from
Jun 20, 2024
Merged

Fix unfixed naming inconsistencies #531

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
2 changes: 1 addition & 1 deletion modyn/tests/common/grpc/test_grpc_helpers.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@ def test_prepare_start_training_request(
assert req.seed == 0
assert req.num_samples_to_pass == (num_samples_to_pass if num_samples_to_pass is not None else 0)
assert req.shuffle
assert req.measure_operation_time
assert req.enable_accurate_gpu_measurements


@patch("modyn.common.grpc.grpc_helpers.grpc_connection_established", return_value=True)
Expand Down
22 changes: 10 additions & 12 deletions modyn/trainer_server/internal/trainer/pytorch_trainer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -115,7 +115,7 @@
self._device = device
self._device_type = "cuda" if "cuda" in self._device else "cpu"
self._amp = training_info.amp
self._measure_operation_time = training_info.measure_operation_time
self._measure_gpu_ops = training_info.enable_accurate_gpu_measurements
self._checkpoint_path = training_info.checkpoint_path
self._checkpoint_interval = training_info.checkpoint_interval
self._final_checkpoint_path = training_info.final_checkpoint_path
Expand Down Expand Up @@ -241,7 +241,7 @@

if self._sample_then_batch_this_epoch(epoch):
self.update_queue("TRAINING", batch_number, self._num_samples, training_active=False)
with GPUMeasurement(self._measure_operation_time, "DownsampleSTB", self._device, stopw):
with GPUMeasurement(self._measure_gpu_ops, "DownsampleSTB", self._device, stopw):

Check warning on line 244 in modyn/trainer_server/internal/trainer/pytorch_trainer.py

View check run for this annotation

Codecov / codecov/patch

modyn/trainer_server/internal/trainer/pytorch_trainer.py#L244 

Added line #L244 was not covered by tests
self.downsample_trigger_training_set()

stopw.start("IndivFetchBatch", overwrite=True)
Expand All @@ -258,7 +258,7 @@

self.update_queue("TRAINING", batch_number, self._num_samples, training_active=True)

with GPUMeasurement(self._measure_operation_time, "PreprocessBatch", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "PreprocessBatch", self._device, stopw, resume=True):
sample_ids, target, data = self.preprocess_batch(batch, stopw)

if retrieve_weights_from_dataloader:
Expand All @@ -271,9 +271,7 @@

with torch.autocast(self._device_type, enabled=self._amp):
if self._downsampling_mode == DownsamplingMode.BATCH_THEN_SAMPLE:
with GPUMeasurement(
self._measure_operation_time, "DownsampleBTS", self._device, stopw, resume=True
):
with GPUMeasurement(self._measure_gpu_ops, "DownsampleBTS", self._device, stopw, resume=True):
data, sample_ids, target, weights = self.downsample_batch(data, sample_ids, target)
self._assert_data_size(post_downsampling_size, data, sample_ids, target)
if not batch_accumulator.inform_batch(data, sample_ids, target, weights):
Expand All @@ -284,10 +282,10 @@
data, sample_ids, target, weights = batch_accumulator.get_accumulated_batch()

self._assert_data_size(self._batch_size, data, sample_ids, target)
with GPUMeasurement(self._measure_operation_time, "Forward", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "Forward", self._device, stopw, resume=True):
output = self._model.model(data)

with GPUMeasurement(self._measure_operation_time, "Loss", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "Loss", self._device, stopw, resume=True):
if weighted_optimization:
# weighted gradient descent
assert weights is not None
Expand All @@ -302,10 +300,10 @@
)
stopw.stop()

with GPUMeasurement(self._measure_operation_time, "Backward", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "Backward", self._device, stopw, resume=True):
self._scaler.scale(loss).backward()

with GPUMeasurement(self._measure_operation_time, "OptimizerStep", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "OptimizerStep", self._device, stopw, resume=True):
for _, optimizer in self._optimizers.items():
self._scaler.step(optimizer)

Expand Down Expand Up @@ -505,10 +503,10 @@
target = batch[2]
stopw.stop("LabelTransform")

with GPUMeasurement(self._measure_operation_time, "MoveLabelToGPU", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "MoveLabelToGPU", self._device, stopw, resume=True):
target = target.to(self._device)

with GPUMeasurement(self._measure_operation_time, "MoveDataToGPU", self._device, stopw, resume=True):
with GPUMeasurement(self._measure_gpu_ops, "MoveDataToGPU", self._device, stopw, resume=True):
data: Union[torch.Tensor, dict]
if isinstance(batch[1], torch.Tensor):
data = batch[1].to(self._device)
Expand Down
Loading