model-evaluator 1.0.13

Library for beautiful training loop for PyTorch

Model Evaluator

Description

A package for training and evaluating PyTorch models.

Now it supports only visualization of training of models, but in future will support training and evaluation of many models which can be very useful for students.

If you have any bugs/improvement ideas, create an issue thread.

GitHub repository

PyPi

Installation

For installation you need to execute following command:

pip install model-evaluator

Then you can import library and use it in your code:

import model_evaluator

Example usage

from model_evaluator import Trainer, PlottingOptions, Accuracy, F1Score

train_data_loader = ...
test_data_loader = ...
model = ...
optimizer = ...
criterion = ... # those are yours

trainer = Trainer(model, optimizer, criterion, [Accuracy(), F1Score()],
                  plotting_options=PlottingOptions.PLOT_BOTH, plot_interval=8)
trainer.train(NUM_EPOCHS, train_data_loader, test_data_loader, device)

Authors

Artsiom Halian - GitHub

Versions

1.0.13 - Added cpu loss for weighted losses
1.0.12 - Fixes
1.0.11 - Fixes
1.0.10 - Fixes
1.0.9 - Fixes
1.0.8 - Fixes
1.0.7 - Added HuggingFace support
1.0.6 - Explicit conversion to tensor
1.0.5 - Fixed plotting
1.0.4 - Fixed binary recognition
1.0.3 - Fixed multiclass backward error
1.0.2 - Fixed history error
1.0.1 - Updated readme
1.0.0 - Initial version