AlexNet b

Click a block to read its description, or select it with Tab and Enter.

AlexNet bClassification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.LibreYOLOAlexNet bClassification, 224 × 224 input, 1,000 classes. Tensor sizes exclude batch.Features 1Input3 × 224 × 224Conv2d 11×11 / 464 × 55 × 55; p=2, g=1ReLU64 × 55 × 55MaxPool2d64 × 27 × 27Conv2d 5×5 / 1192 × 27 × 27; p=2, g=1ReLU192 × 27 × 27MaxPool2d192 × 13 × 13Features 2Conv2d 3×3 / 1384 × 13 × 13; p=1, g=1ReLU384 × 13 × 13Conv2d 3×3 / 1256 × 13 × 13; p=1, g=1ReLU256 × 13 × 13Conv2d 3×3 / 1256 × 13 × 13; p=1, g=1ReLU256 × 13 × 13MaxPool2d256 × 6 × 6ClassifierAdaptiveAvgPool2d256 × 6 × 6Flatten9,216Dropout (eval identity)p=0.5; 9216Linear9216 to 4096ReLU4096Dropout (eval identity)p=0.5; 4096Linear4096 to 4096ReLU4096Linear4096 to 1000Native single-tower graph: no local response normalization or grouped convolution.Convolutions and Linear layers include bias. MaxPool kernels are 3×3 with stride 2.Output: 1,000 logits. Dropout is inactive in the displayed eval graph.Source: libreyolo/models/alexnet/nn.py. Revision a4d0ecc9e17f.libreyolo.com