Skip to content

latent_dataset: stride support in random mode, accept passed filters #485

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

Draft
ElmoPA wants to merge 1 commit into
base: elmo/eval-latent-knn
Choose a base branch
from
Draft

latent_dataset: stride support in random mode, accept passed filters #485

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
39 changes: 32 additions & 7 deletions egomimic/eval/latent_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -61,6 +61,11 @@
"_shuffle_random",
"_shuffle_pairs",
"_shuffle_custom",
"use_tokenizer",
"model_name",
"sampling_mode",
"annotation_key",
"default_prompt",
}
)

Expand All @@ -86,18 +91,38 @@ def build_dataset(
swallowing) — typos in yaml fail loudly.
"""
if mode == "random":
lam = (
"lambda row: row['task'] == "
f"{task!r} and row['robot_name'] == {embodiment!r}"
)
filters = DatasetFilter(filter_lambdas=[lam])
logger.info("[build_dataset] %s | random mode | filter=%s", embodiment, lam)
return MultiDataset._from_resolver(
if filters is not None:
logger.info(
"[build_dataset] %s | random mode | using passed filters", embodiment
)
else:
lam = (
"lambda row: row['task'] == "
f"{task!r} and row['robot_name'] == {embodiment!r}"
)
filters = DatasetFilter(filter_lambdas=[lam])
logger.info("[build_dataset] %s | random mode | filter=%s", embodiment, lam)
base = MultiDataset._from_resolver(
resolver,
filters=filters,
mode="total",
valid_ratio=valid_ratio,
)
if stride is not None and stride > 0:
logger.info(
"[build_dataset] %s | random+stride=%d wrapping",
embodiment,
stride,
)
return EvenStrideDataset(base, stride=stride)
if frames_per_episode is not None:
logger.info(
"[build_dataset] %s | random+frames_per_episode=%d wrapping",
embodiment,
frames_per_episode,
)
return EvenStrideDataset(base, frames_per_episode=frames_per_episode)
return base

if mode in ("pairs", "custom"):
if not hashes:
Expand Down
Loading