import pyarrow.parquet as pq | |
class TatoebaLarge: | |
def __init__(self, file_path): | |
self.file_path = file_path | |
self.parquet_file = pq.ParquetFile(self.file_path) | |
def __iter__(self): | |
return self.parquet_file.iter_batches() | |
def __len__(self): | |
return self.parquet_file.metadata.num_rows | |
# Usage example | |
dataset = ParquetDataset("path/to/your/dataset.parquet") | |
for batch in dataset: | |
for example in batch.to_pydict(): | |
# Do something with the example, e.g., preprocess the data | |
print(example) |