Struct BurnGraph 

pub struct BurnGraph { /* private fields */ }

Burn graph intermediate representation of modules and tensor operations.

Implementations

impl BurnGraph

pub fn register(&mut self, node: Node)

Register a new operation node into the graph.

Notes

The node must be registered in the same order they will be executed in the forward pass.

pub fn with_burnpack( self, out_file: PathBuf, strategy: LoadStrategy, ) -> BurnGraph

Save the state of each node in a burnpack file and generate weight-loading constructors.

The LoadStrategy controls which constructors are generated on the Model struct.

pub fn with_blank_space(self, blank_spaces: bool) -> BurnGraph

Add blank spaces in some places

Notes

It can be problematic when testing.

pub fn with_top_comment(self, top_comment: Option<String>) -> BurnGraph

Add a comment at the top of the generated file.

pub fn with_partition(self, partition: bool) -> BurnGraph

Enable or disable submodule partitioning for large models.

pub fn codegen(self) -> TokenStream

Generate tokens representing the graph with Burn modules and tensor operations.

pub fn register_input_output( &mut self, input_names: Vec<String>, output_names: Vec<String>, input_args: &[Argument], output_args: &[Argument], )

Register the input and output types of the graph using the passed in names. The names must be unique and match the names of the inputs and outputs of the nodes. The order will be preserved.

Arguments

• input_names - The names of the inputs of the graph.
• output_names - The names of the outputs of the graph.
• input_args - The input arguments (from ONNX graph, used for empty graphs).
• output_args - The output arguments (from ONNX graph, used for empty graphs).

Trait Implementations

impl Debug for BurnGraph

fn fmt(&self, f: &mut Formatter<'_>) -> Result<(), Error>

Formats the value using the given formatter.

impl Default for BurnGraph

fn default() -> BurnGraph

Returns the “default value” for a type.