liveBook · Manning

Ray Data

As explained in Transformers in Action

Ray Data is a scalable data-processing library for machine-learning workloads within the Ray ecosystem.

Ray Data is a scalable data-processing library for machine-learning workloads. It supplies Ray-native datasets that support distributed preprocessing and training while retaining compatibility with Hugging Face datasets. section 8.2

SuperGLUE CommitmentBank pipeline

Convert the CommitmentBank train, validation, and test splits into Ray datasets, then feed training and validation datasets into a distributed Transformer classification workflow.

How Transformers in Action covers it

How deep it goes
Demonstrates dataset conversion, sharding, batch iteration, and trainer integration, but does not detail Ray Data internals.
Examples
Uses SuperGLUE CommitmentBank splits, Ray-native dataset conversion, tokenized batches, and Transformer classification training.
What it assumes
Assumes familiarity with Hugging Face datasets and PyTorch-style model training while avoiding required distributed-systems expertise.
What you can do after
Apply Ray Data via Ray-native datasets in distributed preprocessing, training, evaluation, and tuning pipelines.
How it connects
Connects Hugging Face datasets with tokenization, PyTorch training, Ray Tune, and distributed hyperparameter optimization.
Cover of Transformers in Action
liveBookmanning.com
Read Transformers in Action on liveBook Get the book at manning.com

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Transformers in Action. The section links point to the parts of the book it draws on.