head.py

# Copyright 2020 The AutoKeras Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Optional

import keras

from autokeras.engine import io_hypermodel
from autokeras.utils import types


def serialize_metrics(metrics):
    serialized = []
    for metric in metrics:
        if isinstance(metric, str):
            serialized.append([metric])
        else:
            serialized.append(keras.metrics.serialize(metric))
    return serialized


def deserialize_metrics(metrics):
    deserialized = []
    for metric in metrics:
        if isinstance(metric, list):
            deserialized.append(metric[0])
        else:
            deserialized.append(keras.metrics.deserialize(metric))
    return deserialized


def serialize_loss(loss):
    if isinstance(loss, str):
        return [loss]
    return keras.losses.serialize(loss)


def deserialize_loss(loss):
    if isinstance(loss, list):
        return loss[0]
    return keras.losses.deserialize(loss)


class Head(io_hypermodel.IOHyperModel):
    """Base class for the heads, e.g. classification, regression.

    # Arguments
        loss: A Keras loss function. Defaults to None. If None, the loss will be
            inferred from the AutoModel.
        metrics: A list of Keras metrics. Defaults to None. If None, the metrics
            will be inferred from the AutoModel.
    """

    def __init__(
        self,
        loss: Optional[types.LossType] = None,
        metrics: Optional[types.MetricsType] = None,
        **kwargs
    ):
        super().__init__(**kwargs)
        self.loss = loss
        if metrics is None:
            metrics = []
        self.metrics = metrics
        # Mark if the head should directly output the input tensor.

    def get_config(self):
        config = super().get_config()
        config.update(
            {
                "loss": serialize_loss(self.loss),
                "metrics": serialize_metrics(self.metrics),
            }
        )
        return config

    @classmethod
    def from_config(cls, config):
        config["loss"] = deserialize_loss(config["loss"])
        config["metrics"] = deserialize_metrics(config["metrics"])
        return super().from_config(config)

    def build(self, hp, inputs=None):
        raise NotImplementedError