def run_mnist_prediction(prediction_hparams):
# DATA
datamodule = MNISTDataModule(data_dir=DATA_PATH + '/mnist_')
datamodule.prepare_data() # downloads data to given path
train_dataloader = datamodule.train_dataloader()
val_dataloader = datamodule.val_dataloader()
test_dataloader = datamodule.test_dataloader()
encoder_model = SimpleConvNet()
return run_prediction(prediction_hparams, train_dataloader, val_dataloader, test_dataloader,
encoder_model, encoder_out_dim=50)
def run_mnist_dvrl(prediction_hparams, dvrl_hparams):
# DATA
datamodule = MNISTDataModule(data_dir=DATA_PATH + '/mnist_')
datamodule.prepare_data() # downloads data to given path
train_dataloader = datamodule.train_dataloader(
batch_size=dvrl_hparams.get('outer_batch_size', 32))
val_dataloader = datamodule.val_dataloader(
batch_size=dvrl_hparams.get('outer_batch_size', 32))
test_dataloader = datamodule.test_dataloader(
batch_size=dvrl_hparams.get('outer_batch_size', 32))
val_split = datamodule.val_split
encoder_model = SimpleConvNet()
return run_gumbel(dvrl_hparams,
prediction_hparams,
train_dataloader,
val_dataloader,
test_dataloader,
val_split,
encoder_model,
encoder_out_dim=50)
loss = fn.cross_entropy(y_pred, y_true)
loss.backward()
siamese_net_opt.step()
net.train()
siamese_net.eval()
net_opt.zero_grad()
y_pred, _ = net(x, siamese_net)
loss = fn.cross_entropy(y_pred, y_true)
train_loss.append(loss.detach().cpu().numpy())
accuracy = (torch.argmax(y_pred, 1) == y_true).float().mean()
train_accuracy.append(accuracy.detach().cpu().numpy())
loss.backward()
net_opt.step()
print(f"{eid} | Train: loss={np.mean(train_loss):.6f}, acc={np.mean(train_accuracy):.4f}")
net.eval()
val_accuracy, val_loss = [], []
for batch in mnist_dm.val_dataloader(val_batch_size):
x, y_true = batch[0].cuda(device_id), batch[1].cuda(device_id)
y_pred, _ = net(x, siamese_net)
loss = fn.cross_entropy(y_pred, y_true)
val_loss.append(loss.detach().cpu().numpy())
accuracy = (torch.argmax(y_pred, 1) == y_true).float().mean()
val_accuracy.append(accuracy.detach().cpu().numpy())
print(f"{eid} | Val: loss={np.mean(val_loss):.6f}, acc={np.mean(val_accuracy):.4f}")
device_id = None
latest_checkpoint = ""
if __name__ == '__main__':
train_batch_size = 1024
val_batch_size = 2048
data_dir = '../data/'
train_transforms = None
val_transforms = None
mnist_dm = MNISTDataModule(data_dir,
train_transforms=train_transforms,
val_transforms=val_transforms)
mlp = DeepELM(1)
train_pipe = TrainPipe(model=mlp)
classic_trainer = pl.Trainer(
gpus=([device_id] if device_id is not None else None), max_epochs=1000)
print("TRAINING MLP")
classic_trainer.fit(
train_pipe,
train_dataloader=mnist_dm.train_dataloader(train_batch_size),
val_dataloaders=mnist_dm.val_dataloader(val_batch_size))
# print("SAVE MLP")
# torch.save(train_pipe.state_dict(), 'model_v1')
# print("TEST MLP")
# classic_trainer.test(train_pipe, datamodule=mnist_dm)