Looks like dynamically setting input size fails for Quantum LSTM

[DanielAtKrypton]

Description of the problem

If you look at the logs:

tests/early_stopping_test.py ssss
tests/main_test.py sweight_shapes = (n_qlayers, n_qubits) = (1, 6)
Using device cpu
Re-initializing module because the following parameters were re-set: input_dim, output_dim.
weight_shapes = (n_qlayers, n_qubits) = (1, 2)
Re-initializing optimizer.
  epoch    train_loss    cp      dur
-------  ------------  ----  -------
      1        0.1551     +  11.5193

I was expecting that the reported weight_shapes still be (1, 6) after parameters re-set as specified in the test code.

[BenjaminBossan]

I had a look at your code but couldn't follow what happened. My best guess is that the shape is re-set to 2 because that's the default as defined here. If you want to always have it at 6, maybe you can specify that directly as a grid search parameter (with only the value 6).

[DanielAtKrypton]

I had a look at your code but couldn't follow what happened. My best guess is that the shape is re-set to 2 because that's the default as defined here. If you want to always have it at 6, maybe you can specify that directly as a grid search parameter (with only the value 6).

It was supposed to work as documented here. Since the code is using InputShapeSetter.

[BenjaminBossan]

I still cannot really figure out what happens in your code, which test is called with what data, where weight_shapes = (n_qlayers, n_qubits) = (1, 2) is printed etc. Ideally, you could produce a minimal code example that reproduces the error.

[BenjaminBossan]

@DanielAtKrypton any updates?

[DanielAtKrypton]

@DanielAtKrypton any updates?

I will come back to this and follow your suggestion of minimal code example as soon I have a couple higher priorities sorted out. Thank you @BenjaminBossan.