Partial Training

OrtValueCache

class onnxruntime.capi._pybind_state.OrtValueCache(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache) → None

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache) → None

clear(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache) → None

count(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache, arg0: str) → int

insert(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache, arg0: str, arg1: onnxruntime.capi.onnxruntime_pybind11_state.OrtValue) → None

keys(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache) → list

remove(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache, arg0: str) → None

TrainingAgent

class onnxruntime.capi._pybind_state.TrainingAgent(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingAgent, arg0: onnxruntime.capi.onnxruntime_pybind11_state.InferenceSession, arg1: List[str], arg2: List[onnxruntime.capi.onnxruntime_pybind11_state.OrtDevice], arg3: List[str], arg4: List[onnxruntime.capi.onnxruntime_pybind11_state.OrtDevice]) → None

This is the main class used to run a ORTModule model.

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingAgent, arg0: onnxruntime.capi.onnxruntime_pybind11_state.InferenceSession, arg1: List[str], arg2: List[onnxruntime.capi.onnxruntime_pybind11_state.OrtDevice], arg3: List[str], arg4: List[onnxruntime.capi.onnxruntime_pybind11_state.OrtDevice]) → None

run_backward(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingAgent, arg0: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueVector, arg1: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueVector, arg2: onnxruntime.capi.onnxruntime_pybind11_state.PartialGraphExecutionState) → None

run_forward(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingAgent, arg0: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueVector, arg1: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueVector, arg2: onnxruntime.capi.onnxruntime_pybind11_state.PartialGraphExecutionState, arg3: onnxruntime.capi.onnxruntime_pybind11_state.OrtValueCache) → None

PartialGraphExecutionState

class onnxruntime.capi._pybind_state.PartialGraphExecutionState(self: onnxruntime.capi.onnxruntime_pybind11_state.PartialGraphExecutionState) → None

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.PartialGraphExecutionState) → None

OrtModuleGraphBuilder

class onnxruntime.capi._pybind_state.OrtModuleGraphBuilder(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) → None

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) → None

build(*args, **kwargs)

Overloaded function.

1. build(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) -> None

2. build(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder, arg0: List[List[int]]) -> None

get_graph_info(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) → onnxruntime.capi.onnxruntime_pybind11_state.GraphInfo

get_inference_optimized_model(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) → bytes

get_model(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder) → bytes

initialize(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilder, arg0: bytes, arg1: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilderConfiguration) → None

OrtModuleGraphBuilderConfiguration

class onnxruntime.capi._pybind_state.OrtModuleGraphBuilderConfiguration(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilderConfiguration) → None

Configuration information for module graph builder.

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.OrtModuleGraphBuilderConfiguration) → None

property build_gradient_graph

property enable_caching

property graph_transformer_config

property initializer_names

property initializer_names_to_train

property input_names_require_grad

property loglevel

property use_memory_efficient_gradient

TrainingGraphTransformerConfiguration

class onnxruntime.capi._pybind_state.TrainingGraphTransformerConfiguration(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingGraphTransformerConfiguration) → None

Training Graph transformer configuration.

__init__(self: onnxruntime.capi.onnxruntime_pybind11_state.TrainingGraphTransformerConfiguration) → None

property allow_layer_norm_mod_precision

property attn_dropout_recompute

property enable_gelu_approximation

property gelu_recompute

property number_recompute_layers

property propagate_cast_ops_config

property transformer_layer_recompute

previous

Training

next

Training with onnxruntime and pytorch