Petastorm

Petastorm is a library developed by Uber for deep learning model training and evaluation. It supports single-machine and distributed workflows with datasets in Apache Parquet format, is compatible with TensorFlow, PyTorch, and PySpark, and can be used via pure Python code.

GitHub repo Directory import
0 shouts Added Sep 23, 2026 github.com
github.com/uber/petastorm
Petastorm website preview

Links

Tags