lovinervy/AMT-Apple
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

switch info to debug in logger

Browse Source
This commit is contained in:
Viner Abubakirov
2026-03-31 10:11:18 +05:00
parent cf9f0350ce
commit 89e795e320
1 changed files with 17 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
main.py
View File

@@ -84,44 +84,44 @@ class ImageInterpolator:
self.vram_available = get_vram_available(device) self.vram_available = get_vram_available(device)
self.embt = torch.tensor(1 / 2).float().view(1, 1, 1, 1).to(device) self.embt = torch.tensor(1 / 2).float().view(1, 1, 1, 1).to(device)
self.model_runner = model_runner self.model_runner = model_runner
logging.info( logging.debug(
f"Initialized ImageInterpolator with device: {device}, anchor: {anchor}, available VRAM: {self.vram_available} bytes" f"Initialized ImageInterpolator with device: {device}, anchor: {anchor}, available VRAM: {self.vram_available} bytes"
) )
def interpolate(self, image1: Path, image2: Path, output_path: Path): def interpolate(self, image1: Path, image2: Path, output_path: Path):
logging.info(f"Reading images: {image1} and {image2}") logging.debug(f"Reading images: {image1} and {image2}")
tensor1 = img2tensor(utils.read(image1)).to(self.device) tensor1 = img2tensor(utils.read(image1)).to(self.device)
tensor2 = img2tensor(utils.read(image2)).to(self.device) tensor2 = img2tensor(utils.read(image2)).to(self.device)
logging.info( logging.debug(
f"Image shapes after conversion to tensors: {tensor1.shape}, {tensor2.shape}" f"Image shapes after conversion to tensors: {tensor1.shape}, {tensor2.shape}"
) )
tensor1, tensor2 = check_dim_and_resize(tensor1, tensor2) tensor1, tensor2 = check_dim_and_resize(tensor1, tensor2)
logging.info(f"Image shapes after resizing: {tensor1.shape}, {tensor2.shape}") logging.debug(f"Image shapes after resizing: {tensor1.shape}, {tensor2.shape}")
h, w = tensor1.shape[2], tensor1.shape[3] h, w = tensor1.shape[2], tensor1.shape[3]
logging.info(f"Interpolating images of size: {h}x{w}") logging.debug(f"Interpolating images of size: {h}x{w}")
scale = self.scale(h, w) scale = self.scale(h, w)
logging.info(f"Calculated scale factor: {scale:.2f}") logging.debug(f"Calculated scale factor: {scale:.2f}")
padding = int(16 / scale) padding = int(16 / scale)
logging.info(f"Calculated padding: {padding} pixels") logging.debug(f"Calculated padding: {padding} pixels")
padder = InputPadder(tensor1.shape, divisor=padding) padder = InputPadder(tensor1.shape, divisor=padding)
tensor1_padded, tensor2_padded = padder.pad(tensor1, tensor2) tensor1_padded, tensor2_padded = padder.pad(tensor1, tensor2)
logging.info( logging.debug(
f"Image shapes after padding: {tensor1_padded.shape}, {tensor2_padded.shape}" f"Image shapes after padding: {tensor1_padded.shape}, {tensor2_padded.shape}"
) )
tensor1_padded = tensor1_padded.to(self.device) tensor1_padded = tensor1_padded.to(self.device)
tensor2_padded = tensor2_padded.to(self.device) tensor2_padded = tensor2_padded.to(self.device)
logging.info("Running model inference for interpolation") logging.debug("Running model inference for interpolation")
with torch.no_grad(): with torch.no_grad():
interpolated = self.model_runner.model( interpolated = self.model_runner.model(
tensor1_padded, tensor2_padded, self.embt, scale_factor=scale, eval=True tensor1_padded, tensor2_padded, self.embt, scale_factor=scale, eval=True
)["imgt_pred"] )["imgt_pred"]
logging.info(f"Interpolated image shape before unpadding: {interpolated.shape}") logging.debug(f"Interpolated image shape before unpadding: {interpolated.shape}")
(interpolated,) = padder.unpad(interpolated) (interpolated,) = padder.unpad(interpolated)
logging.info(f"Interpolated image shape after unpadding: {interpolated.shape}") logging.debug(f"Interpolated image shape after unpadding: {interpolated.shape}")
utils.write(output_path, tensor2img(interpolated.cpu())) utils.write(output_path, tensor2img(interpolated.cpu()))
logging.info(f"Saved interpolated image to: {output_path}") logging.debug(f"Saved interpolated image to: {output_path}")
def scale(self, height: int, width: int) -> float: def scale(self, height: int, width: int) -> float:
scale = ( scale = (
@@ -145,7 +145,9 @@ def main():
ckpt_path = Path("src/pretrained/amt-g.pth") ckpt_path = Path("src/pretrained/amt-g.pth")
image1_path = Path("source/img0.png") image1_path = Path("source/img0.png")
image2_path = Path("source/img1.png") image2_path = Path("source/img1.png")
output_path = Path("output/interpolated_image.png") image3_path = Path("source/img2.png")
output_path1 = Path("output/interpolated_image1.png")
output_path2 = Path("output/interpolated_image2.png")
device = get_device() device = get_device()
model_runner = ModelRunner(config_path, ckpt_path, device) model_runner = ModelRunner(config_path, ckpt_path, device)
@@ -166,7 +168,8 @@ def main():
else: else:
raise Exception(f"Unsupported device type: {device.type}") raise Exception(f"Unsupported device type: {device.type}")
interpolator = ImageInterpolator(device, anchor, model_runner) interpolator = ImageInterpolator(device, anchor, model_runner)
interpolator.interpolate(image1_path, image2_path, output_path) interpolator.interpolate(image1_path, image2_path, output_path1)
interpolator.interpolate(image2_path, image3_path, output_path2)
if __name__ == "__main__": if __name__ == "__main__":