def test_loss_DANN(model_name, pretrained):
""" Loss and model should work together without exceptions"""
# for pretrained in [True, False]:
dann_config.MODEL_BACKBONE = model_name
dann_config.BACKBONE_PRETRAINED = pretrained
if model_name == 'vanilla_dann' and pretrained:
return
model = DANNModel()
batch = random_batch(unknown_proportion=0.5)
loss_DANN(model, batch, 7, 130)
batch = random_batch(unknown_proportion=1.00)
loss_DANN(model, batch, 7, 130)
batch = random_batch(unknown_proportion=0.0)
loss_DANN(model, batch, 7, 130)
def __init__(self, metric):
self.metric = metric
def reset(self):
self.metric.reset()
def __call__(self, *args, **kwargs):
pass
print(f"Call metric with args:\n{args}\n{kwargs}")
print("metric: ", self.metric(*args, **kwargs))
if __name__ == '__main__':
train_gen_s, val_gen_s, test_gen_s = create_data_generators(
"office-31", 'amazon', batch_size=16, infinite_train=True)
train_gen_t, val_gen_t, test_gen_t = create_data_generators(
"office-31", 'dslr', batch_size=16, infinite_train=True)
model = DANNModel()
acc = AccuracyScoreFromLogits()
mmm = DebugMetric(acc)
tr = Trainer(model, only_loss)
tr.fit(train_gen_s,
train_gen_t,
n_epochs=5,
validation_data=[val_gen_s, val_gen_t],
metrics=[acc],
steps_per_epoch=1,
callback=simple_callback)
batch_size=dann_config.BATCH_SIZE,
infinite_train=True,
image_size=dann_config.IMAGE_SIZE,
num_workers=dann_config.NUM_WORKERS,
device=device)
train_gen_t, val_gen_t, test_gen_t = create_data_generators(
dann_config.DATASET,
dann_config.TARGET_DOMAIN,
batch_size=dann_config.BATCH_SIZE,
infinite_train=True,
image_size=dann_config.IMAGE_SIZE,
num_workers=dann_config.NUM_WORKERS,
device=device)
model = DANNModel().to(device)
acc = AccuracyScoreFromLogits()
mmm = DebugMetric(acc)
scheduler = LRSchedulerSGD()
tr = Trainer(model, loss_DANN)
tr.fit(train_gen_s,
train_gen_t,
n_epochs=dann_config.N_EPOCHS,
validation_data=[val_gen_s, val_gen_t],
metrics=[acc],
steps_per_epoch=dann_config.STEPS_PER_EPOCH,
val_freq=dann_config.VAL_FREQ,
opt='sgd',
opt_kwargs={
'lr': 0.01,
parser.add_argument('--checkpoint',
type=str,
required=True,
help='path to model checkpoint')
args = parser.parse_args()
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
test_gen_t, _, _ = create_data_generators(
dann_config.DATASET,
dann_config.TARGET_DOMAIN,
batch_size=dann_config.BATCH_SIZE,
infinite_train=False,
image_size=dann_config.IMAGE_SIZE,
split_ratios=[1, 0, 0],
num_workers=dann_config.NUM_WORKERS,
device=device)
model = DANNModel().to(device)
model.load_state_dict(torch.load(args.checkpoint))
model.eval()
acc = AccuracyScoreFromLogits()
tr = Trainer(model, None)
scores = tr.score(test_gen_t, [acc])
scores_string = ' '.join(
['{}: {:.5f}\t'.format(k, float(v)) for k, v in scores.items()])
print(
f"scores on dataset \"{dann_config.DATASET}\", domain \"{dann_config.TARGET_DOMAIN}\":"
)
print(scores_string)