Skip to content

dgisolfi/belt

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

9 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

belt

A composable machine learning toolbelt

Pipelines

ML Tasks are broken into pipelines allowing models to be used with multiple pipelines and datasets.

Pipeline Model Dataset
classifier SoftmaxRegressor (NumPy) Iris
translation Seq2Seq RNN/LSTM multi30k, opus_books

Usage

# Softmax classifier
python -m belt classifier

Configs live in configs/. Each pipeline has a YAML controlling data, model, training, and output settings.

Extending

New model: register with the decorator, then set model_name in your config:

from belt.supervised.models import supervised_model_registry

@supervised_model_registry.register("NewModel")
class NewModel(nn.Module):
    pass

New dataset: add a loader to belt/data/ and register it in belt/registry.py.

Data cache

Downloaded datasets and tokenizers are stored under data/ at the repo root and reused on subsequent runs. Pass --no-cache to force a fresh download.

Install

python -m pip install -e .

# DirectML backend (Windows GPU):
python -m pip install -e .[directml]

About

A composable machine learning toolbet

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages