config4ml

Basic configuration utilities and API for ML projects

NOTE: This repo is in pre-alpha and under development!!!

Change-log

[v0.0.1]

• [data] Transform config & registry
• [data] Dataset base config
• [data] MNIST built-in config (also as tutorial)
• [lightning] Lightning trainer config with callbacks support

[v0.0.2]

• [lightning] Support for Neptune, Tensorboard logger configs
• [lightning] Support for Console logging

[v0.0.3]

• [lightning] Trainer with console logging override option
• [lightning] More arguments for Early Stopping callback
• [lightning] Weights and biases logger (Wand) support
• [torch] Support for torch LR schedulers config [stepLR, exponentialLR, reduceOnPlateauLR]
• [torch] Support for torch Optimizers config [Adam, SGD]

[v0.0.3b]

• [lightning] Trainer support for "resume_from_checkpoint" argument (default: None)

[v0.0.3c]

• [data] Transform config separately for train / val