HeyTensor

HeyTensor is a browser based tensor shape calculator for people building and debugging neural networks. It computes the output shape of a layer, shows the formula behind the result, and helps you find which layer parameter is causing a PyTorch shape error before you rerun your training code. The calculator covers 14 layer types across PyTorch and TensorFlow, including Conv2d, Conv1d, Linear, LSTM, GRU, MultiheadAttention, BatchNorm, pooling layers, Flatten, Reshape and Concatenate. Chain Mode lets you stack layers and watch the shape at every step, highlighting the point where two layers become incompatible. Paste Error Mode accepts common PyTorch RuntimeError messages, pulls out the dimension numbers and suggests which parameters to change. Architecture presets such as LeNet 5, a simple CNN, a ResNet block, a Transformer encoder layer and an LSTM classifier give you a starting point. It suits machine learning students, researchers and engineers who spend time on shape mismatches such as a Linear layer expecting the wrong number of input features. The site also offers Tensor Preflight, a downloadable PyTorch toolkit that traces a local forward pass, produces an HTML report of tensor shapes and turns a fix into a repeatable regression check. It requires Python 3.9 or later and PyTorch 2.8 or later and runs locally. The calculator is free and runs in your browser. Tensor Preflight is a one time payment of $29 for a ZIP download that includes source code, setup instructions and eight broken and fixed example workflows, and a sample report is available to view before buying.