SymbolicModel

pytorch_symbolic.SymbolicModel

pytorch_symbolic.SymbolicModel(inputs: Tuple[SymbolicData, ...] | List[SymbolicData] | SymbolicData, outputs: Tuple[SymbolicData, ...] | List[SymbolicData] | SymbolicData, enable_cuda_graphs = False, enable_forward_codegen = None)

Bases: nn.Module

A PyTorch model that replays operations defined in the graph.

All operations that were required to change inputs into outputs will be replayed in the same order, but on the real data provided as input to this model.

Example::

input1 = Input((10,))
input2 = Input((10,))
x = input1 + input2
x = nn.Linear(x.features, 1)(x)
model = SymbolicModel((input1, input2), x)

Parameters:

Name	Type	Description	Default
`inputs`	`Tuple[SymbolicData, ...] \| List[SymbolicData] \| SymbolicData`	A collection of SymbolicData that represent the input data used by the model. It is you who provide the specific data when the model is created. If you have mulitple inputs here, be prepared to pass multiple inputs during training/inference.	required
`outputs`	`Tuple[SymbolicData, ...] \| List[SymbolicData] \| SymbolicData`	A collection of SymbolicTensors that will end the computations. These nodes return your final computation result. So if you have mulitple outputs, SymbolicModel will return a tuple of tensors.	required
`enable_cuda_graphs`		If True, after the model creation, model will be converted to CUDA Graph. This requires CUDA capable device. CUDA Graphs are greatly speeding up the execution of some of the models. Not all models are compatible with CUDA Graphs. For example, if your model includes non-deterministic behaviour, it likely won't work.	`False`

Attributes:

Name	Type	Description
`inputs`	`tuple`	Non-modifiable tuple of input nodes
`outputs`	`tuple`	Non-modifiable tuple of output nodes

input_shape `property`

input_shape

Return shape of the input or in case of multiple inputs - a tuple of them.

output_shape `property`

output_shape

Return shape of the output or in case of multiple outputs - a tuple of them.

forward

forward(inputs: torch.Tensor) -> Any

This function is executed by call. Do not use this directly, use call instead.

Warning!

This function will be overwritten by _replace_forward_with_codegen if enable_forward_codegen is True. If this happened and you want to see your source, print self._generated_forward_source.

summary

summary()

Print Keras-like model summary.

deepcopy

__deepcopy__(memo)

This copies a working Module, but the underlying graph structure is not copied!