torchmanager 1.1

PyTorch Training Manager v1.1

torchmanager

A highly-wrapped PyTorch training and testing manager

This branch is used for beta unstable release. Please check main branch for the latest main release version

Pre-request

• Python 3.8+
• PyTorch
• tqdm
• tensorboard (Optional)

Installation

• PyPi: pip install torchmanager-nightly
• Conda: conda install -c kisonho torchmanager-nightly

The Manager

• Initialize the manager with target model, optimizer, loss function, and metrics:

import torch, torchmanager

# define model
class PytorchModel(torch.nn.Module):
    ...

# initialize model, optimizer, loss function, and metrics
model = PytorchModel(...)
optimizer = torch.optim.SGD(model.parameters())
loss_fn = torchmanager.losses.CrossEntropy()
metrics = {'accuracy': torchmanager.metrics.SparseCategoricalAccuracy()}

# initialize manager
manager = torchmanager.Manager(model, optimizer, loss_fn=loss_fn, metrics=metrics)

• Train the model with fit method:

from torchmanager.data import Dataset

# get datasets
training_dataset: Dataset = ...
val_dataset: Dataset = ...

# train with fit method
manager.fit(training_dataset, epochs=10, val_dataset=val_dataset)

• Test the model with test method:

# get dataset
testing_dataset: Dataset = ...

# test with test method
manager.test(testing_dataset)

• There are also some other callbacks to use:

...

tensorboard_callback = torchmanager.callbacks.TensorBoard('logs')
last_ckpt_callback = torchmanager.callbacks.LastCheckpoint(model, 'last.model')
manager.fit(..., callbacks_list=[tensorboard_callback, last_ckpt_callback])

Custom your training loop

1. Create your own manager class by extending the Manager class:

...

class CustomManager(Manager):
    ...

2. Override the train_step method:

class CustomManager(Manager):
    ...
    
    def train_step(x_train: torch.Tensor, y_train: torch.Tensor) -> Dict[str, float]:
        ...