New Model failed to work on micro speech

I have just completed training a new model on the Micro Speech. I have the same input size (1960) and preprocessed using the Microfrontend. but when I deployed on arduino. I get this error.

Requested feature_data_ size 536907080 doesn’t match 1960
Feature generation failed.

Is there any solution to this. The model works fine on TFLITE in python but always throw this error on serial Monitor.

Hmm. Did you see this issue, it looks the same:

Yes I have seen this and other suggested solutions online but they didn’t work. I’m not just sure what is going on. I’ve even tried to use older version of tensorflow to train and convert the model into Lite version but didn’t work