spyrit.core.train

Training functions for deep learning models.

Functions

attr_removal(old_key, old_name)
attr_transformation(old_key, old_name, new_name)
boxplot(model1, model2, model3, criterion, ...)
boxplotconsist(model1, model2, model3, ...)
checkpoint(root, epoch, model)	Saves the dictionaries of a given pytorch model for the right epoch
compare_model(model1, model2, model3, ...[, ...])	Compare three models
count_memory(model)
count_param(model)
count_trainable_param(model)
images_norm(images)
imshow(img[, title])
load_net(title, model[, device, strict])	Loads net defined by title
multiplot(train_info1, train_info2, train_info3)
read_param(path)
remove_model_attributes(source, old_name[, ...])	Remove some attributes of a saved model (nn.module)
rename_model_attributes(source, old_name, ...)	Rename the name of the attributes of a saved model (nn.module)
save_net(title, model)	Saves dictionaries of a given pytorch model in the place defined by title
tb_profiler(path_prof, model, criterion, ...)	Tensorboard profiler: Profile code execution
tb_writer_add_image(writer, name_metric, ...)	Tensorboard writer: Add an image)
tb_writer_add_scalar(writer, name_metric, ...)	Tensorboard writer: Add a scalar (loss)
tb_writer_init(tb_path[, samples])	Tensorboard log for torch
train_model(model, criterion, optimizer, ...)	Trains the pytorch model
train_model_supervised(model, criterion, ...)	Trains the pytorch model in a supervised way
visualize_conv_layers(conv_layer[, ...])	Displays the 8 first filters of the convolution layer conv_layer
visualize_model(model, dataloaders, device)	Takes 8 images from the dataloader and shows side by side the input image and the reconstructed image

Classes

Train_par(batch_size, learning_rate, img_size)
Weight_Decay_Loss(loss)