BlackHC/toma

Helps you write algorithms in PyTorch that adapt to the available (CUDA) memory

44
/ 100
Emerging

When running deep learning models in PyTorch, you often encounter 'out-of-memory' errors, especially when using GPUs. This tool automatically adjusts the size of the data batches your model processes, or the chunks it operates on, to fit within the available memory. It takes your PyTorch code and, if it fails due to memory limits, retries with smaller data batches until it succeeds. Machine learning engineers and researchers who train or run inference on large models will find this useful.

437 stars. Used by 1 other package. No commits in the last 6 months. Available on PyPI.

Use this if you are frequently encountering CUDA out-of-memory errors when running PyTorch models and want an automated way to adapt your batch or chunk sizes.

Not ideal if your operations are not memory-intensive, as there is a small overhead involved in the memory adaptation process.

deep-learning GPU-optimization pytorch model-training inference-optimization
Stale 6m
Maintenance 0 / 25
Adoption 11 / 25
Maturity 25 / 25
Community 8 / 25

How are scores calculated?

Stars

437

Forks

10

Language

Python

License

MIT

Last pushed

Aug 29, 2024

Commits (30d)

0

Dependencies

2

Reverse dependents

1

Get this data via API

curl "https://pt-edge.onrender.com/api/v1/quality/ml-frameworks/BlackHC/toma"

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