I am training my model in Slurm and using the preprocessed UW-Madison GI Tract Medical Segmentation Dataset from learnopencv.com.

The Model trained around 82 epochs but at the half of 83 epochs it stops and show this error FileNotFoundError: [Errno 2] No such file or directory: './data/dataset_UWM_GI_Tract_train_valid/train/images/case77_day20_slice_0095_266_266_1.50_1.50.png' .

.

I have used another model to train with same dataset and eventually it works fine.

This is my code for dataset.

class MedicalDataset1(Dataset):
    def __init__(self, image_paths, mask_paths, img_size, ds_mean, ds_std, is_train=False):
        self.image_paths = image_paths
        self.mask_paths  = mask_paths  
        self.is_train    = is_train
        self.img_size    = img_size
        self.ds_mean = ds_mean
        self.ds_std = ds_std
        self.transforms  = self.setup_transforms(mean=self.ds_mean, std=self.ds_std)

        self.train_imgs = sorted(os.listdir(image_paths))
        self.train_msks  = sorted(os.listdir(mask_paths))
 
    def __len__(self):
        return  len(self.train_imgs)
 
    def setup_transforms(self, mean, std):
        transforms = []
 
        if self.is_train:
            transforms.extend([
                
                A.HorizontalFlip(p=0.5), A.VerticalFlip(p=0.5),
                A.ShiftScaleRotate(scale_limit=0.12, rotate_limit=0.15, shift_limit=0.12, p=0.5),
                A.RandomBrightnessContrast(p=0.5),
                A.CoarseDropout(max_holes=8, max_height=self.img_size[1]//20, max_width=self.img_size[0]//20, min_holes=5, fill_value=0, mask_fill_value=0, p=0.5)
            ])
        transforms.extend([
                A.Normalize(mean=mean, std=std, always_apply=True),
                ToTensorV2(always_apply=True),  # (H, W, C) --> (C, H, W)
        ])
        return A.Compose(transforms)

    def __getitem__(self, idx):

        img_filename = self.train_imgs[idx]
        mask_filename = self.train_msks[idx]
        
        img_path = os.path.join(self.image_paths, img_filename)
        mask_path = os.path.join(self.mask_paths, mask_filename)
        
    
        image = np.array(Image.open(img_path).convert('RGB').resize((224,224),Image.NEAREST))
        mask = np.array(Image.open(mask_path).resize((224,224),Image.NEAREST))
         
        transformed = self.transforms(image=image, mask=mask)
        image, mask = transformed["image"], transformed["mask"].to(torch.long)
        return image, mask