Skip to content

[feat] use parquet.ParquetFile instead of pa.Dataset to prevent ParquetDataset OOM #125

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

Merged
tiankongdeguiji merged 1 commit into from
Mar 4, 2025
Merged

[feat] use parquet.ParquetFile instead of pa.Dataset to prevent ParquetDataset OOM #125

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
22 changes: 14 additions & 8 deletions tzrec/datasets/parquet_dataset.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,7 +18,6 @@
from typing import Any, Dict, Iterator, List, Optional

import pyarrow as pa
import pyarrow.dataset as ds
from pyarrow import parquet

from tzrec.constant import Mode
Expand All @@ -27,6 +26,16 @@
from tzrec.protos import data_pb2


def _reader_iter(
input_files: List[str],
batch_size: int,
) -> Iterator[pa.RecordBatch]:
for input_file in input_files:
parquet_file = parquet.ParquetFile(input_file)
for batch in parquet_file.iter_batches(batch_size):
yield batch


class ParquetDataset(BaseDataset):
"""Dataset for reading data with parquet format.

Expand Down Expand Up @@ -93,16 +102,16 @@ def __init__(
self._input_files.extend(glob.glob(input_path))
if len(self._input_files) == 0:
raise RuntimeError(f"No parquet files exist in {self._input_path}.")
dataset = ds.dataset(self._input_files[0], format="parquet")
parquet_file = parquet.ParquetFile(self._input_files[0])
if self._selected_cols:
self._ordered_cols = []
for field in dataset.schema:
for field in parquet_file.schema_arrow:
# pyre-ignore [58]
if field.name in selected_cols:
self.schema.append(field)
self._ordered_cols.append(field.name)
else:
self.schema = dataset.schema
self.schema = parquet_file.schema_arrow

def to_batches(
self, worker_id: int = 0, num_workers: int = 1
Expand All @@ -112,10 +121,7 @@ def to_batches(
if self._shuffle:
random.shuffle(input_files)
if len(input_files) > 0:
dataset = ds.dataset(input_files, format="parquet")
reader = dataset.to_batches(
batch_size=self._batch_size, columns=self._ordered_cols
)
reader = _reader_iter(input_files, self._batch_size)
yield from self._arrow_reader_iter(reader)

def num_files(self) -> int:
Expand Down