def forward_feature(self, batch, **kwargs):
data = list_to_variable(batch['image_batch'],
volatile=kwargs.get('volatile', False))
labels = tensor_to_var(torch.Tensor(np.array(batch['label_batch'])))
features = self.model.features(data)
return features, labels
def forward(self, batch, **kwargs):
"""
:param batch:
:param kwargs:
:return:
"""
# print(batch['image_batch'][0].min())
# print(batch['image_batch'][0].max())
data = list_to_variable(batch['image_batch'],
volatile=kwargs.get('volatile', False))
lbls = [img / 255.0 for img in batch['image_batch']]
labels = list_to_variable(lbls, volatile=kwargs.get('volatile', False))
pred = self.model(data)
return pred, labels
def forward(self, batch, **kwargs):
"""
:param batch:
:param kwargs:
:return:
"""
data = list_to_variable(batch['image_batch'],
volatile=kwargs.get('volatile', False))
labels = tensor_to_var(torch.Tensor(np.array(batch['label_batch'])))
pred = self.model(data)
return pred, labels
def forward(self, batch, **kwargs):
"""
:param batch:
:param kwargs:
:return:
"""
data = list_to_variable(batch['image_batch'],
volatile=kwargs.get('volatile', False))
labels_lesion = tensor_to_var(
torch.Tensor(np.array(batch['label_batch'])))
labels_row = tensor_to_var(
torch.Tensor(np.array(batch['center_row_batch'])))
labels_col = tensor_to_var(
torch.Tensor(np.array(batch['center_col_batch'])))
pred = self.model(data)
return pred[:,
0], pred[:,
1], pred[:,
2], labels_lesion, labels_row, labels_col
def test_resnet_auto():
model = get_feature_model('auto',
'auto',
load_pretrained=False,
opti=None,
lr=None,
mom=None,
checkpoint_pretrained=None,
learn_pos=False,
force_on_cpu=False)
val_loader = get_valloader_only(False, False, 1, 0, False)
for batch_idx, batch in enumerate(val_loader):
if len(batch['image_batch']) > 0:
# print(len(batch['image_batch']))
# print(batch['image_batch'][0].size())
data = list_to_variable(batch['image_batch'])
labels = tensor_to_var(torch.Tensor(np.array(
batch['label_batch'])),
async=True)
pred = model(data)
break