Extensor.NIF (Extensor v2.3.2)

NIF wrapper module for tensorflow adapter functions

Link to this section Summary

Functions

module initialization callback

loads a custom op kernel library

loads a saved_model from a path into a new tensorflow session

loads a graph_def protobuf into a new tensorflow session

executes the graph in a running session

Link to this section Types

Link to this type

tensor_map()

Specs

tensor_map() :: %{required(String.t()) => {integer(), tuple(), binary()}}

Link to this section Functions

Specs

init() :: :ok

module initialization callback

Link to this function

tf_load_library(name)

Specs

tf_load_library(name :: String.t()) :: :ok

loads a custom op kernel library

Link to this function

tf_load_saved_model(path, tag, config_pb)

Specs

tf_load_saved_model(
  path :: String.t(),
  tag :: String.t(),
  config_pb :: binary()
) :: {reference(), binary()}

loads a saved_model from a path into a new tensorflow session

Link to this function

tf_parse_frozen_graph(graph_pb, config_pb)

Specs

tf_parse_frozen_graph(graph_pb :: binary(), config_pb :: binary()) ::
  reference()

loads a graph_def protobuf into a new tensorflow session

Link to this function

tf_run_session(session, input_tensors, output_names)

Specs

tf_run_session(
  session :: reference(),
  input_tensors :: tensor_map(),
  output_names :: [String.t()]
) :: tensor_map()

executes the graph in a running session