tfc.ops.entropy_encode_channel

Encodes each input in value.

tfc.ops.entropy_encode_channel(
    handle, value, name=None
)

In general, entropy encoders in handle reference multiple distributions. The last (innermost) dimension of value determines which distribution is used to encode value. For example, if value is a 3-D array, then value(i,j,k) is encoded using the k-th distribution.

handle controls the number of coding streams. Suppose that value has the shape [2, 3, 4] and that handle has shape [2]. Then the first slice [0, :, :] of shape [3, 4] is encoded into handle[0] and the second slice [1, :, :] is encoded into handle[1]. If handle has shape [], then there is only one handle, and the entire input is encoded into a single stream.

Values must be in the provided ranges specified when the input handle was originally created, unless overflow functionality was enabled. The handle may be produced by the CreateRangeEncoder op, or may be passed through from a different EntropyEncodeChannel/EntropyEncodeIndex op.

Because the op modifies handle, the corresponding input edge to the op nodes of this type should not have other consumers in the graph.