Skip to main content
E0010: input dimension mismatch

​
What happened

You passed a spike raster with got input features to a model that expects expected input features. The dimensions do not match and the model cannot run.

​
Why

The model was compiled for a specific input size (e.g. 700 for SHD). The input you provided has a different number of features.

​
How to fix

  1. Check your encoder output shape. The raster shape should be [T, n_features] where n_features matches the first layer’s in_features.
  2. If you are using thrindex.encoders, verify the n_channels parameter matches the model architecture.
  3. If you are building input manually, check that you are not transposing T and N dimensions.

​
Example

​
Common mistake

MNIST has 784 pixels. SHD has 700 cochlear channels. If you accidentally pass an MNIST-encoded input to an SHD model, you get this error.