Crate tract

Structs

• TractFact
• TractInferenceFact
• TractInferenceModel
• TractModel
• TractNnef
• TractOnnx
• TractRunnable
• TractState
• TractValue

Enums

• TRACT_RESULT
  Used as a return type of functions that can encounter errors. If the function encountered an error, you can retrieve it using the tract_get_last_error function

Functions

• tract_fact_destroy^⚠
• tract_fact_dump^⚠
  Write a fact as its specification string.
• tract_fact_parse^⚠
  Parse a fact specification string into an Fact.
• tract_free_cstring^⚠
  Frees a string allocated by libtract.
• tract_get_last_error
  Retrieve the last error that happened in this thread. A function encountered an error if its return type is of type TRACT_RESULT and it returned TRACT_RESULT_KO.
• tract_inference_fact_destroy^⚠
  Destroy a fact.
• tract_inference_fact_dump^⚠
  Write an inference fact as its specification string.
• tract_inference_fact_empty^⚠
  Creates an empty inference fact.
• tract_inference_fact_parse^⚠
  Parse a fact specification string into an InferenceFact.
• tract_inference_model_analyse^⚠
  Analyse an InferencedModel in-place.
• tract_inference_model_destroy^⚠
  Destroy an InferenceModel.
• tract_inference_model_input_count^⚠
  Query an InferenceModel input counts.
• tract_inference_model_input_fact^⚠
• tract_inference_model_input_name^⚠
  Query the name of a model input.
• tract_inference_model_into_optimized^⚠
  Convenience function to obtain an optimized TypedModel from an InferenceModel.
• tract_inference_model_into_typed^⚠
  Transform a fully analysed InferenceModel to a TypedModel.
• tract_inference_model_output_count^⚠
  Query an InferenceModel output counts.
• tract_inference_model_output_fact^⚠
  Query an output fact for an InferenceModel.
• tract_inference_model_output_name^⚠
  Query the name of a model output.
• tract_inference_model_set_input_fact^⚠
  Set an input fact of an InferenceModel.
• tract_inference_model_set_output_fact^⚠
  Set an output fact of an InferenceModel.
• tract_inference_model_set_output_names^⚠
  Change the model outputs nodes (by name).
• tract_model_concretize_symbols^⚠
  Give value one or more symbols used in the model.
• tract_model_declutter^⚠
  Declutter a TypedModel in-place.
• tract_model_destroy^⚠
  Destroy a TypedModel.
• tract_model_input_count^⚠
  Query an InferenceModel input counts.
• tract_model_input_fact^⚠
  Query the input fact of a model.
• tract_model_input_name^⚠
  Query the name of a model input.
• tract_model_into_runnable^⚠
  Convert a TypedModel into a TypedRunnableModel.
• tract_model_optimize^⚠
  Optimize a TypedModel in-place.
• tract_model_output_count^⚠
  Query an InferenceModel output counts.
• tract_model_output_fact^⚠
  Query the output fact of a model.
• tract_model_output_name^⚠
  Query the name of a model output.
• tract_model_profile_json^⚠
  Perform a profile of the model using the provided inputs.
• tract_model_property^⚠
  Query a property value in a model.
• tract_model_property_count^⚠
  Query the number of properties in a model.
• tract_model_property_names^⚠
  Query the properties names of a model.
• tract_model_pulse_simple^⚠
  Pulsify the model
• tract_model_set_output_names^⚠
  Change the model outputs nodes (by name).
• tract_model_transform^⚠
  Apply a transform to the model.
• tract_nnef_create^⚠
  Creates an instance of an NNEF framework and parser that can be used to load and dump NNEF models.
• tract_nnef_destroy^⚠
  Destroy the NNEF parser. It is safe to detroy the NNEF parser once the model had been loaded.
• tract_nnef_enable_extended_identifier_syntax^⚠
• tract_nnef_enable_onnx^⚠
• tract_nnef_enable_pulse^⚠
• tract_nnef_enable_tract_core^⚠
• tract_nnef_enable_tract_extra^⚠
• tract_nnef_model_for_path^⚠
  Parse and load an NNEF model as a tract TypedModel.
• tract_nnef_write_model_to_dir^⚠
  Dump a TypedModel as a NNEF directory.
• tract_nnef_write_model_to_tar^⚠
  Dump a TypedModel as a NNEF tar file.
• tract_nnef_write_model_to_tar_gz^⚠
  Dump a TypedModel as a NNEF .tar.gz file.
• tract_onnx_create^⚠
  Creates an instance of an ONNX framework and parser that can be used to load models.
• tract_onnx_destroy^⚠
  Destroy the NNEF parser. It is safe to detroy the NNEF parser once the model had been loaded.
• tract_onnx_model_for_path^⚠
  Parse and load an ONNX model as a tract InferenceModel.
• tract_runnable_input_count^⚠
  Query a Runnable input counts.
• tract_runnable_output_count^⚠
  Query an Runnable output counts.
• tract_runnable_release^⚠
• tract_runnable_run^⚠
  Convenience function to run a stateless model.
• tract_runnable_spawn_state^⚠
  Spawn a session state from a runnable model.
• tract_state_destroy^⚠
• tract_state_input_count^⚠
  Query a State input counts.
• tract_state_output_count^⚠
  Query an State output counts.
• tract_state_run^⚠
  Run a turn on a model state
• tract_value_as_bytes^⚠
  Inspect part of a value. Except value, all argument pointers can be null if only some specific bits are required.
• tract_value_destroy^⚠
  Destroy a value.
• tract_value_from_bytes^⚠
  Create a TractValue (aka tensor) from caller data and metadata.
• tract_version
  Returns a pointer to a static buffer containing a null-terminated version string.