It creates an input in the network: a place where data, such as features and labels, should be provided.
op_input_variable(shape, dtype = "float32", needs_gradient = FALSE, is_sparse = FALSE, dynamic_axes = c(get_default_batch_axis()), name = "")
shape | - list of ints representing tensor shape integer vector for dimensions of input tensor |
---|---|
needs_gradient | logical whether to conduct backprop on the tensor |
is_sparse | logical whether variable is sparse |
dynamic_axes | list of dynamic axis (only a single axis can be dynamic, i.e., either batch axis or time axis) |
Variable https://www.cntk.ai/pythondocs/cntk.variables.html#cntk.variables.Variable
https://www.cntk.ai/pythondocs/cntk.ops.html#cntk.ops.input_variable