Skip to content

Fix chunk casting and schema unification in dataset #7873

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
wants to merge 2 commits into
base: main
Choose a base branch
from
+53 −3
Open

Fix chunk casting and schema unification in dataset #7873

Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
fa8ba23
Fix chunk casting and schema unification in dataset
ArjunJagdale Nov 19, 2025
a12c854
Update iterable_dataset.py
ArjunJagdale Nov 19, 2025
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
56 changes: 53 additions & 3 deletions src/datasets/iterable_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -564,7 +564,25 @@ def _iter_arrow(self) -> Iterator[tuple[Key, pa.Table]]:
self._state_dict["batch_idx"] += 1
self._state_dict["num_chunks_since_previous_state"] += len(chunks_buffer)
self._state_dict["cropped_chunk_length"] = 0
yield new_key, pa.Table.from_batches(chunks_buffer)

if self.features:
expected_schema = pa.schema(self.features.type)
casted_chunks = []
for chunk in chunks_buffer:
try:
casted_chunks.append(chunk.cast(expected_schema))
except (pa.ArrowInvalid, pa.ArrowNotImplementedError):
casted_chunks.append(chunk)
yield new_key, pa.Table.from_batches(casted_chunks)
else:
# Unify schemas when no explicit features provided
if chunks_buffer:
unified_schema = pa.unify_schemas([chunk.schema for chunk in chunks_buffer])
casted_chunks = [chunk.cast(unified_schema) for chunk in chunks_buffer]
yield new_key, pa.Table.from_batches(casted_chunks, schema=unified_schema)
else:
yield new_key, pa.Table.from_batches(chunks_buffer)

keys_buffer = []
chunks_buffer = []
chunks_buffer_size = 0
Expand All @@ -580,7 +598,23 @@ def _iter_arrow(self) -> Iterator[tuple[Key, pa.Table]]:
self._state_dict["batch_idx"] += 1
self._state_dict["num_chunks_since_previous_state"] += len(chunks_buffer)
self._state_dict["cropped_chunk_length"] = cropped_chunk_length
yield new_key, pa.Table.from_batches(chunks_buffer)

if self.features:
expected_schema = pa.schema(self.features.type)
casted_chunks = []
for chunk in chunks_buffer:
try:
casted_chunks.append(chunk.cast(expected_schema))
except (pa.ArrowInvalid, pa.ArrowNotImplementedError):
casted_chunks.append(chunk)
yield new_key, pa.Table.from_batches(casted_chunks)
else:
if chunks_buffer:
unified_schema = pa.unify_schemas([chunk.schema for chunk in chunks_buffer])
casted_chunks = [chunk.cast(unified_schema) for chunk in chunks_buffer]
yield new_key, pa.Table.from_batches(casted_chunks, schema=unified_schema)
else:
yield new_key, pa.Table.from_batches(chunks_buffer)
keys_buffer = [f"{key}[{cropped_chunk_length}:]"]
chunks_buffer = [chunk.slice(cropped_chunk_length, len(chunk) - cropped_chunk_length)]
chunks_buffer_size = len(chunk) - cropped_chunk_length
Expand All @@ -596,7 +630,23 @@ def _iter_arrow(self) -> Iterator[tuple[Key, pa.Table]]:
self._state_dict["batch_idx"] += 1
self._state_dict["num_chunks_since_previous_state"] = 0
self._state_dict["cropped_chunk_length"] = 0
yield new_key, pa.Table.from_batches(chunks_buffer)

if self.features:
expected_schema = pa.schema(self.features.type)
casted_chunks = []
for chunk in chunks_buffer:
try:
casted_chunks.append(chunk.cast(expected_schema))
except (pa.ArrowInvalid, pa.ArrowNotImplementedError):
casted_chunks.append(chunk)
yield new_key, pa.Table.from_batches(casted_chunks)
else:
if chunks_buffer:
unified_schema = pa.unify_schemas([chunk.schema for chunk in chunks_buffer])
casted_chunks = [chunk.cast(unified_schema) for chunk in chunks_buffer]
yield new_key, pa.Table.from_batches(casted_chunks, schema=unified_schema)
else:
yield new_key, pa.Table.from_batches(chunks_buffer)

def shuffle_data_sources(self, generator: np.random.Generator) -> "RebatchedArrowExamplesIterable":
return RebatchedArrowExamplesIterable(
Expand Down