kir-gadjello/zipslicer

A library for incremental loading of large PyTorch checkpoints

38
/ 100
Emerging

This tool helps machine learning engineers and researchers manage very large PyTorch models. It allows you to selectively load parts of a saved model checkpoint, rather than the entire file at once. This means you can inspect or work with specific layers or parameters without needing enough RAM to hold the whole model, which is especially useful for massive deep learning models.

No commits in the last 6 months. Available on PyPI.

Use this if you are working with extremely large PyTorch model checkpoints and your machine's memory struggles to load the entire model for analysis or further processing.

Not ideal if your application always requires loading the full model's parameters immediately, or if you are working with general-purpose Python objects saved as pickles instead of standard PyTorch state dictionaries.

deep-learning-research large-model-deployment neural-network-analysis resource-optimization model-checkpointing
Stale 6m
Maintenance 0 / 25
Adoption 8 / 25
Maturity 25 / 25
Community 5 / 25

How are scores calculated?

Stars

56

Forks

2

Language

Python

License

Last pushed

Mar 03, 2023

Commits (30d)

0

Dependencies

1

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/kir-gadjello/zipslicer"

Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.