def train(model_params, **training_kwargs):
model_config = make_config(**model_params)
training_params = vars(parse_args())
training_params.update(training_kwargs)
hparams = dict(model_config=model_config)
hparams.update(training_params)
model = SCAEMNIST(Namespace(**hparams))
print("___________")
print("loading model")
# model = SCAEMNIST()
# model = SCAEMNIST.load_from_checkpoint('/content/lightning_logs/version_0/checkpoints/epoch=48.ckpt')
# training_params.update(resume_from_checkpoint='/content/lightning_logs/version_10/checkpoints/epoch=72.ckpt')
checkpoint_callback = ModelCheckpoint(save_top_k=1)
training_params.update(checkpoint_callback=checkpoint_callback)
# if 'save_top_k' in training_params:
# checkpoint_callback = ModelCheckpoint(
# save_top_k=training_params['save_top_k'])
# training_params.update(checkpoint_callback=checkpoint_callback)
# del training_params['save_top_k']
training_params.update(gpus=1, auto_select_gpus=True)
trainer = Trainer(**training_params)
trainer.fit(model)
def get_mdl_obj():
from torch_scae_experiments.wdata.hparams import model_params
model_config = make_config(**model_params)
hparams = dict(model_config=model_config)
model = SCAEMNIST(Namespace(**hparams))
return model
def test_scae(self):
config = Namespace(**factory.make_config(**model_params))
cnn_encoder = CNNEncoder(**config.pcae_cnn_encoder)
part_encoder = CapsuleImageEncoder(encoder=cnn_encoder,
**config.pcae_encoder)
template_generator = TemplateGenerator(
**config.pcae_template_generator)
part_decoder = TemplateBasedImageDecoder(**config.pcae_decoder)
obj_encoder = SetTransformer(**config.ocae_encoder_set_transformer)
obj_decoder_capsule = CapsuleLayer(**config.ocae_decoder_capsule)
obj_decoder = CapsuleObjectDecoder(obj_decoder_capsule)
scae = SCAE(part_encoder=part_encoder,
template_generator=template_generator,
part_decoder=part_decoder,
obj_encoder=obj_encoder,
obj_decoder=obj_decoder,
**config.scae)
with torch.no_grad():
batch_size = 24
image = torch.rand(batch_size, *config.image_shape)
label = torch.randint(0, config.n_classes, (batch_size, ))
reconstruction_target = image
res = scae(image=image)
loss = scae.loss(res, reconstruction_target, label)
accuracy = scae.calculate_accuracy(res, label)
def train(model_params, **training_kwargs):
model_config = make_config(**model_params)
training_params = vars(parse_args())
training_params.update(training_kwargs)
hparams = dict(model_config=model_config)
hparams.update(training_params)
model = SCAECIFAR10(Namespace(**hparams))
if 'save_top_k' in training_params:
checkpoint_callback = ModelCheckpoint(
save_top_k=training_params['save_top_k'])
training_params.update(checkpoint_callback=checkpoint_callback)
del training_params['save_top_k']
trainer = Trainer(**training_params)
trainer.fit(model)
def train(model_params, **training_kwargs):
model_config = make_config(**model_params)
training_params = vars(parse_args())
training_params.update(training_kwargs)
hparams = dict(model_config=model_config)
hparams.update(training_params)
model = SCAEBONEAGE(Namespace(**hparams))
if 'save_top_k' in training_params:
if ('using_colab' in training_params):
del training_params['using_colab']
checkpoint_callback = ModelCheckpoint(
filepath="/content/drive/My Drive/data/boneage",
save_top_k=training_params['save_top_k'])
else:
checkpoint_callback = ModelCheckpoint(
save_top_k=training_params['save_top_k'])
training_params.update(checkpoint_callback=checkpoint_callback)
del training_params['save_top_k']
trainer = Trainer(**training_params)
trainer.fit(model)
from torch_scae_experiments.mnist.train import SCAEMNIST
from torch_scae_experiments.mnist.hparams import model_params
from torch_scae.factory import make_config
from argparse import Namespace
from torch.utils.data import DataLoader, random_split
from torchvision.datasets import MNIST
from torch import cuda
import pathlib
if cuda.is_available():
cuda.set_device(0)
print(cuda.current_device())
#Model Initialization
model_config = make_config(**model_params)
training_hparams = dict(data_dir=str(pathlib.Path.home() / 'torch-datasets'),
gpus=1,
batch_size=16,
num_workers=0,
max_epochs=100,
learning_rate=1e-4,
optimizer_type='RMSprop',
use_lr_scheduler=True,
lr_scheduler_decay_rate=0.997,
model_config=model_config)
scaemnist = SCAEMNIST(Namespace(**training_hparams))
#Data preparation
data_dir = training_hparams['data_dir']
# train and validation datasets
mnist_train = MNIST(data_dir,