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

Jump to bottom

[NeMo-UX] Match nemo 1's default behavior for drop_last and pad_samples_to_global_batch_size #9707

Merged
merged 5 commits into from
Jul 16, 2024
Merged

[NeMo-UX] Match nemo 1's default behavior for drop_last and pad_samples_to_global_batch_size #9707

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
6d94980
match nemo 1's default behavior for drop_last and pad_samples_to_glob…
ashors1 Jul 12, 2024
91d4dd2
Merge branch 'r2.0.0rc1' into ashors/nemo-ux-dataloading-defaults
marcromeyn Jul 12, 2024
5739de8
Merge branch 'r2.0.0rc1' into ashors/nemo-ux-dataloading-defaults
ashors1 Jul 12, 2024
4f262aa
Merge branch 'r2.0.0rc1' into ashors/nemo-ux-dataloading-defaults
ashors1 Jul 15, 2024
1765559
Merge branch 'r2.0.0rc1' into ashors/nemo-ux-dataloading-defaults
ashors1 Jul 16, 2024
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
9 changes: 6 additions & 3 deletions nemo/lightning/data.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,10 +85,13 @@ def add_megatron_sampler(
rampup_batch_size: Optional[List[int]] = None,
consumed_samples: int = 0,
dataloader_type: Literal["single", "cyclic"] = "single",
drop_last: bool = True,
pad_samples_to_global_batch_size: bool = False,
# data_sharding: bool = False
) -> DataLoader:
from megatron.core import parallel_state

## TODO: expose drop_last and pad_samples_to_global_batch_size args
if dataloader_type == 'single':
batch_sampler = MegatronPretrainingSampler(
total_samples=len(dataloader.dataset),
Expand All @@ -98,8 +101,8 @@ def add_megatron_sampler(
rampup_batch_size=rampup_batch_size,
data_parallel_rank=parallel_state.get_data_parallel_rank(),
data_parallel_size=parallel_state.get_data_parallel_world_size(),
drop_last=getattr(dataloader, "_drop_last", False),
pad_samples_to_global_batch_size=getattr(dataloader, "_pad_samples_to_global_batch_size", False),
drop_last=drop_last,
pad_samples_to_global_batch_size=pad_samples_to_global_batch_size,
)
elif dataloader_type == 'cyclic':
batch_sampler = MegatronPretrainingRandomSampler(
Expand All @@ -108,7 +111,7 @@ def add_megatron_sampler(
micro_batch_size=micro_batch_size,
data_parallel_rank=parallel_state.get_data_parallel_rank(),
data_parallel_size=parallel_state.get_data_parallel_world_size(),
pad_samples_to_global_batch_size=getattr(dataloader, "_pad_samples_to_global_batch_size", False),
drop_last=drop_last,
# data_sharding=data_sharding
)
else:
Expand Down
Loading