control_data_package = DataPackage(regular_invaders, AutoEncodeSelect())
run_fac = SimpleRunFac.resume(r'C:\data\runs\489',
co_ord_conv_data_package)
batch_size = 64
epochs = 100
for model, opt, loss_fn, data_package, trainer, tester, run in run_fac:
dev, train, test, selector = data_package.loaders(
batch_size=batch_size)
model.register_forward_hook(input_viewer.view_input)
model.decoder.register_forward_hook(latent_viewer.view_input)
model.register_forward_hook(output_viewer.view_output)
register_tb(run)
for epoch in tqdm(run.for_epochs(epochs), 'epochs', epochs):
#epoch.register_after_hook(write_histogram)
epoch.execute_before(epoch)
trainer.train(model, opt, loss_fn, train, selector, run, epoch)
handles = Handles()
handles += loss_fn.register_hook(tb_test_loss_term)
tester.test(model, loss_fn, test, selector, run, epoch)
handles.remove()
epoch.execute_after(epoch)
run.save()
Params(MseKldLoss),
co_ord_conv_data_package,
run_name='full_v1'))
#run_fac = SimpleRunFac.resume(r'C:\data\runs\549', co_ord_conv_data_package)
batch_size = 64
epochs = 30
for model, opt, loss_fn, data_package, trainer, tester, run in run_fac:
dev, train, test, selector = data_package.loaders(
batch_size=batch_size)
model.register_forward_hook(input_viewer.view_input)
model.decoder.register_forward_hook(latent_viewer.view_input)
model.register_forward_hook(output_viewer.view_output)
register_tb(run, config)
gui = GUIProgressMeter(description='training')
trainer.register_after_hook(gui.update_train)
tester.register_after_hook(gui.update_test)
for epoch in run.for_epochs(epochs):
#epoch.register_after_hook(write_histogram)
epoch.register_after_hook(gui.end_epoch)
epoch.execute_before(epoch)
handles = Handles()
trainer.train(model, opt, loss_fn, dev, selector, run, epoch)
handles += loss_fn.register_hook(TB().tb_test_loss_term)