public RangeDataset(int stop,
int start = 0,
int step = 1,
TF_DataType output_type = TF_DataType.TF_INT64)
{
var start_tensor = tf.convert_to_tensor((long)start);
var step_tensor = tf.convert_to_tensor((long)step);
var stop_tensor = tf.convert_to_tensor((long)stop);
structure = new TensorSpec[] { new TensorSpec(new int[0], dtype: output_type) };
variant_tensor = ops.range_dataset(start_tensor, stop_tensor, step_tensor, output_types, output_shapes);
}
public InputLayer(InputLayerArgs args) :
base(args)
{
this.args = args;
built = true;
SupportsMasking = true;
if (BatchInputShape != null)
{
args.BatchSize = BatchInputShape.dims[0];
args.InputShape = BatchInputShape.dims.Skip(1).ToArray();
}
// moved to base class
if (string.IsNullOrEmpty(args.Name))
{
var prefix = "input";
name = prefix + '_' + keras.backend.get_uid(prefix);
args.Name = name;
}
if (args.DType == TF_DataType.DtInvalid)
{
args.DType = args.InputTensor == null ? tf.float32 : args.InputTensor.dtype;
}
if (args.InputTensor == null)
{
if (args.InputShape != null)
{
args.BatchInputShape = new int[] { args.BatchSize }
.Concat(args.InputShape.dims)
.ToArray();
}
else
{
args.BatchInputShape = null;
}
var graph = keras.backend.get_graph();
graph.as_default();
args.InputTensor = keras.backend.placeholder(
shape: BatchInputShape,
dtype: DType,
name: Name,
sparse: args.Sparse,
ragged: args.Ragged);
graph.Exit();
isPlaceholder = true;
}
// Create an input node to add to self.outbound_node
// and set output_tensors' _keras_history.
// input_tensor._keras_history = base_layer.KerasHistory(self, 0, 0)
// input_tensor._keras_mask = None
var node = new Node(new NodeArgs
{
Outputs = args.InputTensor
});
node.Connect(this);
typeSpec = new TensorSpec(args.InputTensor.TensorShape,
dtype: args.InputTensor.dtype,
name: Name);
}