Logdir : logs/FancyCNN_3 ## Command /home/fix_jer/GIT/Cours/deeplearning-lectures/LabsSolutions/pytorch/introlab/src/introlab/main.py train config.yaml Config : {'data': {'root_dir': './data', 'batch_size': 32, 'num_workers': 4, 'valid_ratio': 0.2, 'normalize': True}, 'optim': {'algo': 'SGD', 'params': {'lr': 0.01}}, 'nepochs': 20, 'loss': 'CrossEntropyLoss', 'logging': {'logdir': './logs'}, 'model': {'class': 'FancyCNN', 'num_layers': 4}} ## Summary of the model architecture ========================================================================================== Layer (type:depth-idx) Output Shape Param # ========================================================================================== FancyCNN [32, 10] -- ├─Sequential: 1-1 [32, 10] -- │ └─Sequential: 2-1 [32, 256, 1, 1] -- │ │ └─Conv2d: 3-1 [32, 16, 28, 28] 160 │ │ └─ReLU: 3-2 [32, 16, 28, 28] -- │ │ └─BatchNorm2d: 3-3 [32, 16, 28, 28] 32 │ │ └─Conv2d: 3-4 [32, 16, 28, 28] 2,320 │ │ └─ReLU: 3-5 [32, 16, 28, 28] -- │ │ └─BatchNorm2d: 3-6 [32, 16, 28, 28] 32 │ │ └─Conv2d: 3-7 [32, 32, 14, 14] 2,080 │ │ └─ReLU: 3-8 [32, 32, 14, 14] -- │ │ └─BatchNorm2d: 3-9 [32, 32, 14, 14] 64 │ │ └─Conv2d: 3-10 [32, 32, 14, 14] 9,248 │ │ └─ReLU: 3-11 [32, 32, 14, 14] -- │ │ └─BatchNorm2d: 3-12 [32, 32, 14, 14] 64 │ │ └─Conv2d: 3-13 [32, 32, 14, 14] 9,248 │ │ └─ReLU: 3-14 [32, 32, 14, 14] -- │ │ └─BatchNorm2d: 3-15 [32, 32, 14, 14] 64 │ │ └─Conv2d: 3-16 [32, 64, 7, 7] 8,256 │ │ └─ReLU: 3-17 [32, 64, 7, 7] -- │ │ └─BatchNorm2d: 3-18 [32, 64, 7, 7] 128 │ │ └─Conv2d: 3-19 [32, 64, 7, 7] 36,928 │ │ └─ReLU: 3-20 [32, 64, 7, 7] -- │ │ └─BatchNorm2d: 3-21 [32, 64, 7, 7] 128 │ │ └─Conv2d: 3-22 [32, 64, 7, 7] 36,928 │ │ └─ReLU: 3-23 [32, 64, 7, 7] -- │ │ └─BatchNorm2d: 3-24 [32, 64, 7, 7] 128 │ │ └─Conv2d: 3-25 [32, 128, 3, 3] 32,896 │ │ └─ReLU: 3-26 [32, 128, 3, 3] -- │ │ └─BatchNorm2d: 3-27 [32, 128, 3, 3] 256 │ │ └─Conv2d: 3-28 [32, 128, 3, 3] 147,584 │ │ └─ReLU: 3-29 [32, 128, 3, 3] -- │ │ └─BatchNorm2d: 3-30 [32, 128, 3, 3] 256 │ │ └─Conv2d: 3-31 [32, 128, 3, 3] 147,584 │ │ └─ReLU: 3-32 [32, 128, 3, 3] -- │ │ └─BatchNorm2d: 3-33 [32, 128, 3, 3] 256 │ │ └─Conv2d: 3-34 [32, 256, 1, 1] 131,328 │ │ └─ReLU: 3-35 [32, 256, 1, 1] -- │ │ └─BatchNorm2d: 3-36 [32, 256, 1, 1] 512 │ └─AdaptiveAvgPool2d: 2-2 [32, 256, 1, 1] -- │ └─Flatten: 2-3 [32, 256] -- │ └─Linear: 2-4 [32, 10] 2,570 ========================================================================================== Total params: 569,050 Trainable params: 569,050 Non-trainable params: 0 Total mult-adds (Units.MEGABYTES): 418.85 ========================================================================================== Input size (MB): 0.10 Forward/backward pass size (MB): 29.20 Params size (MB): 2.28 Estimated Total Size (MB): 31.58 ========================================================================================== ## Loss CrossEntropyLoss() ## Datasets : Train : WrappedDataset(dataset=, transform=Compose( ToImage() ToDtype(scale=True) RandomHorizontalFlip(p=0.5) RandomRotation(degrees=[-10.0, 10.0], interpolation=nearest, expand=False, fill=0) Normalize(mean=[0.286], std=[0.275], inplace=False) )) Validation : WrappedDataset(dataset=, transform=Compose( ToImage() ToDtype(scale=True) Normalize(mean=[0.286], std=[0.275], inplace=False) ))