Switching the ML Interface

In addition to the torch interface (qibmol.interfaces.pytorch.QuantumModel) showcased in the previous section, a keras based interface is available as well (qibmol.interfaces.keras.QuantumModel). Similarly to torch, this exposes a keras.Model object that can be easily integrated in keras pipelines.

In order to switch between the two interfaces, one can just make use of the corresponding QuantumModel object:

import qiboml.interfaces.pytorch as pt
import qiboml.interfaces.keras as ks

# the torch interface
pt.QuantumModel

# the keras interface
ks.QuantumModel

Note that, since the pre-defined encoding, decoding and ansatz layers provided in qiboml are meant to be interface agnostic, they can be used indiscriminately with either one of the two interfaces:

from qiboml.models.encoding import BinaryEncoding
from qiboml.models.decoding import Probabilities
from qiboml.models.ansatze import hardware_efficient

# these are interface agnostic
encoding = BinaryEncoding(2)
decoding = Probabilities(2)
circuit = hardware_efficient(2)

# build the torch model
torch_model = pt.QuantumModel(
   circuit_structure=[encoding, circuit],
   decoding=decoding,
)
# build the keras model
keras_model = ks.QuantumModel(
   circuit_structure=[encoding, circuit],
   decoding=decoding,
)

As it happens for torch, the keras_model object can then be used in combination with the keras API, e.g. stacking layers through the keras.Sequential or training with the keras.optimizers:

import keras

opt = keras.optimizers.Adam(learning_rate=0.01)
keras_model.compile(loss='categorical_crossentropy', optimizer=opt)
keras_model.fit(...)