torch_to_nnef.torch_graph.ir_data

Abstractions used in torch_to_nnef internal graph data IR.

The goal is that these elements are: - extracted/parsed from PyTorch graph data structs - translated to NNEF graph data structs

Classes:

Name	Description
BlobTorchScriptObject	Used only in Quantized Operators from our current obervation.
Data	Base abstract T2N IR data holder.
DictTensors	Used as transition object only.
FixedTensorList	FixedTensorList is a list that contains tensor constant or not.
TensorDataSlot	Prevent uncontrolled assignation of tensor data in Data nodes.
TensorVariable
TupleTensors	Used as transition object only.

BlobTorchScriptObject dataclass

BlobTorchScriptObject(name: str, data: Any)

Bases: Data

Used only in Quantized Operators from our current obervation.

Data dataclass

Data(name: str, data: Any)

Bases: NamedItem

Base abstract T2N IR data holder.

DictTensors dataclass

DictTensors(name: str, data: Dict[str, TensorVariable])

Bases: Data

Used as transition object only.

None should be remaining once graph is fully expanded

FixedTensorList dataclass

FixedTensorList(name: str, data: Sequence[Union[TensorVariable, PythonConstant]])

Bases: Data

FixedTensorList is a list that contains tensor constant or not.

TensorDataSlot

TensorDataSlot()

Prevent uncontrolled assignation of tensor data in Data nodes.

TensorVariable dataclass

TensorVariable(name: str, data: Any, shape: Optional[List[int]], dtype: Optional[dtype], quant: Optional[Dict[str, Any]] = None, _traced_data: Optional[Tensor] = None, module_attr: bool = False)

Bases: Data

Attributes:

Name	Type	Description
tracing_data		Generate data if is not fixed based on tensor information.

tracing_data property

tracing_data

Generate data if is not fixed based on tensor information.

we use it to produce computation trace

TupleTensors dataclass

TupleTensors(name: str, data: List[TensorVariable])

Bases: Data

Used as transition object only.

None should be remaining once graph is fully expanded