E0109 — DelayLengthMismatch - THRINDEX

E0109: delay array length mismatch

What happened

A Dense layer with delays enabled has a delay array whose length does not equal in_features × out_features. The delay array must have exactly one entry per connection.

Why

The delay array is corrupt in the Graph IR. This should not occur if the model was built with snn.Dense(..., delays=True) using a current version of the Python SDK. If it does occur, there is a serialization bug.

How to fix

Recompile the model from Python: thx.compile(model, "model.thx").
If the error persists across fresh compilations, file a bug report — this indicates a serialization bug in the Python SDK.

Example

E0109: delay array length mismatch in layer 0
Why: a dense delay array must have exactly in_features × out_features = 358400 entries, but it has 358399 entries.
Fix: recompile the model — the Graph IR delay data may be corrupt.
Docs: https://docs.thrindex.com/errors/E0109