Skip to content

Make deterministic an option. #4

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

Open
kevinzakka wants to merge 1 commit into
base: main
Choose a base branch
from
Open

Make deterministic an option. #4

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
7 changes: 4 additions & 3 deletions dlimp/dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -57,11 +57,11 @@ def __getattribute__(self, name):
return _wrap(attr, None)
return attr

def _apply_options(self):
def _apply_options(self, deterministic: bool):
"""Applies some default options for performance."""
options = tf.data.Options()
options.autotune.enabled = True
options.deterministic = False
options.deterministic = deterministic
options.experimental_optimization.apply_default_optimizations = True
options.experimental_optimization.map_fusion = True
options.experimental_optimization.map_and_filter_fusion = True
Expand Down Expand Up @@ -130,6 +130,7 @@ def from_rlds(
split: str = "train",
shuffle: bool = True,
num_parallel_reads: int = tf.data.AUTOTUNE,
deterministic: bool = False,
) -> "DLataset":
"""Creates a DLataset from the RLDS format (which is a special case of the TFDS format).

Expand All @@ -150,7 +151,7 @@ def from_rlds(
num_parallel_calls_for_interleave_files=num_parallel_reads,
interleave_cycle_length=num_parallel_reads,
),
)._apply_options()
)._apply_options(deterministic=deterministic)

dataset = dataset.enumerate().traj_map(_broadcast_metadata_rlds)

Expand Down