Closed arnaudstiegler closed 4 years ago
The model has 4 different inputs.
The basic approach using a tensor of 4 arrays was not working:
test_ds = tf.data.Dataset.from_tensor_slices((x0,x1,x2,x3)).batch(32)
model.predict(test_ds)
was throwing a dimension error.
The fix was to provide a list of numpy arrays as the input:
model.predict([x0,x1,x2,x3])
Had a hard time predicting labels using the concatenated model which is a multi-input model. This issue is a way of keeping a record of how to do it