keras的model.compile()的参数

model.compile(optimizer, loss=None, metrics=None, loss_weights=None, 
              sample_weight_mode=None, weighted_metrics=None, target_tensors=None, 
              **kwargs)

Configures the model for training.

# Arguments
    optimizer: String (name of optimizer) or optimizer instance.
        See [optimizers](/optimizers).
    loss: String (name of objective function) or objective function.
        See [losses](/losses).
        If the model has multiple outputs, you can use a different loss
        on each output by passing a dictionary or a list of losses.
        The loss value that will be minimized by the model
        will then be the sum of all individual losses.
    metrics: List of metrics to be evaluated by the model
        during training and testing.
        Typically you will use `metrics=['accuracy']`.
        To specify different metrics for different outputs of a
        multi-output model, you could also pass a dictionary,
        such as `metrics={'output_a': 'accuracy'}`.
    loss_weights: Optional list or dictionary specifying scalar
        coefficients (Python floats) to weight the loss contributions
        of different model outputs.
        The loss value that will be minimized by the model
        will then be the *weighted sum* of all individual losses,
        weighted by the `loss_weights` coefficients.
        If a list, it is expected to have a 1:1 mapping
        to the model's outputs. If a dict, it is expected to map
        output names (strings) to scalar coefficients.
    sample_weight_mode: If you need to do timestep-wise
        sample weighting (2D weights), set this to `"temporal"`.
        `None` defaults to sample-wise weights (1D).
        If the model has multiple outputs, you can use a different
        `sample_weight_mode` on each output by passing a
        dictionary or a list of modes.
    weighted_metrics: List of metrics to be evaluated and weighted
        by sample_weight or class_weight during training and testing.
    target_tensors: By default, Keras will create placeholders for the
        model's target, which will be fed with the target data during
        training. If instead you would like to use your own
        target tensors (in turn, Keras will not expect external
        Numpy data for these targets at training time), you
        can specify them via the `target_tensors` argument. It can be
        a single tensor (for a single-output model), a list of tensors,
        or a dict mapping output names to target tensors.
    **kwargs: When using the Theano/CNTK backends, these arguments
        are passed into `K.function`.
        When using the TensorFlow backend,
        these arguments are passed into `tf.Session.run`.

# Raises
    ValueError: In case of invalid arguments for
        `optimizer`, `loss`, `metrics` or `sample_weight_mode`.