def run():
args = get_args()
config = from_yaml(Config, Path(args.config).read_text())
state_dict = torch.load(args.model)["state_dict"]
state_dict = {
key[6:]: val
for key, val in state_dict.items() if key.startswith("model.")
}
model = EfficientNet.from_name(config.model, num_classes=1)
print(model.load_state_dict(state_dict))
model.eval()
image = cv2.imread(args.target)[:, :, :3]
images = image.transpose(
(2, 0, 1))[None, :, :, :].astype(np.float32) / 255.0
bgr_image = torch.tensor(images)
images = image.transpose(
(2, 0, 1))[None, ::-1, :, :].astype(np.float32) / 255.0
rgb_image = torch.tensor(images)
with torch.no_grad():
output = model.forward(bgr_image)
prediction = ChannelOrder.BGR if output < 0.5 else ChannelOrder.RGB
print(f"BGR image prediction is {prediction}. Output is {output}")
output = model.forward(rgb_image)
prediction = ChannelOrder.BGR if output < 0.5 else ChannelOrder.RGB
print(f"RGB image prediction is {prediction}. Output is {output}")
cv2.imshow("BGR", image)
cv2.imshow("RGB", image[:, :, ::-1])
cv2.waitKey()
def main():
with open('app.yml') as f:
yml = f.read()
cfg = from_yaml(App, yml)
print(cfg)
load_env(cfg, prefix='APP')
print(cfg)
def run():
args = get_args()
config = from_yaml(Config, Path(args.config).read_text())
logger = TestTubeLogger(
save_dir=str(config.save_dir),
name=config.experiment_name,
version=config.version,
)
app = TrainSystem(config)
trainer = Trainer(
min_epochs=50,
max_epochs=config.epochs,
auto_lr_find=True,
auto_scale_batch_size=True,
logger=logger,
auto_select_gpus=True,
gpus=[0],
num_processes=2,
precision=16,
callbacks=[EarlyStopping(monitor="val_loss")],
)
trainer.fit(app)
trainer.test(app)
def main():
with open('swagger.yml') as f:
yaml = f.read()
swagger = from_yaml(Swagger, yaml)
print(swagger)