Rows = neurons, Columns = input features. Edit input values or weights, click Compute to see outputs. Toggle activation to see how nonlinearity affects result.
Layer shape
Tip: click any weight cell to edit it. Use the sliders to change input features (tokens).
Input Vector (features)
Values are shown numerically and via sliders. These represent token embeddings / feature activations for a single token.
Weight Matrix
Click a matrix cell to edit weight. Double-click to cancel edit.
Outputs (pre and post activation)
Summary
Below we show z = W · x (dot product). Activation is applied to z to produce the final neuron output.