layers.rst

Layers

See the networks documentation for more information about how to specify networks.

Default layer: Function with default argument function, so a lambda function is a short-form specification of a simple transformation layer:

Agent.create(
    ...
    policy=dict(network=[
        (lambda x: tf.clip_by_value(x, -1.0, 1.0)),
        ...
    ]),
    ...
)

Dense layers

tensorforce.core.layers.Dense

tensorforce.core.layers.Linear

Convolutional layers

tensorforce.core.layers.Conv1d

tensorforce.core.layers.Conv2d

tensorforce.core.layers.Conv1dTranspose

tensorforce.core.layers.Conv2dTranspose

Embedding layers

tensorforce.core.layers.Embedding

Recurrent layers (unrolled over timesteps)

tensorforce.core.layers.Rnn

tensorforce.core.layers.Lstm

tensorforce.core.layers.Gru

Input recurrent layers (unrolled over sequence input)

tensorforce.core.layers.InputRnn

tensorforce.core.layers.InputLstm

tensorforce.core.layers.InputGru

Pooling layers

tensorforce.core.layers.Flatten

tensorforce.core.layers.Pooling

tensorforce.core.layers.Pool1d

tensorforce.core.layers.Pool2d

Normalization layers

tensorforce.core.layers.LinearNormalization

tensorforce.core.layers.ExponentialNormalization

tensorforce.core.layers.InstanceNormalization

tensorforce.core.layers.BatchNormalization

Misc layers

tensorforce.core.layers.Reshape

tensorforce.core.layers.Activation

tensorforce.core.layers.Dropout

tensorforce.core.layers.Clipping

tensorforce.core.layers.Image

tensorforce.core.layers.Deltafier

tensorforce.core.layers.Sequence

Special layers

tensorforce.core.layers.Function

tensorforce.core.layers.Register

tensorforce.core.layers.Retrieve

tensorforce.core.layers.Block

tensorforce.core.layers.Reuse

Keras layer

tensorforce.core.layers.Keras