Parses an ONNX Gemm operator into a LayerNode. More...

Inheritance diagram for GemmParser:

[legend]

Collaboration diagram for GemmParser:

[legend]

Public Member Functions
LayerNode	run (self)
	Run the parser. More...

dict[str, Any]	get_layer_node_user_format (self, list[int] input_shape, list[int] output_shape)
	Generate layer data in user input format for MatMul or GEMM ONNX nodes. More...

def	generate_layer_node (self)

list[int]	infer_input_activation_shape (self, list[int] output_shape)
	In case the input activations are empty (which can happen if there is a shape operator in the path): extract the weights from the model graph initializer to get the correct input activation size. More...

Public Member Functions inherited from ONNXOperatorParser
None	__init__ (self, int node_id, NodeProto node, dict[int, Any] nodes_outputs, ModelProto onnx_model, *list[dict[str, Any]]\|None mapping_data=None, Accelerator\|None accelerator=None)

def	get_input_output_weight_data_type (self)
	Return the data type of the input, output and weight tensors of this node. More...

def	get_weight_name (self, NodeProto node)
	Return the name of the weight input of this node depending on its operator type. More...

list[int]	get_node_predecessors (self)
	Compute node input sources. More...

def	get_operand_source_user_format (self, list[int] predecessors)
	Set input source and indicate constant operands. More...

def	get_weight_precision (self)
	Return the weight precision for this node. More...

def	get_activation_precision (self)
	Return the activation precision for this node. More...

def	get_intermediate_output_precision (self)
	Return the intermediate output precision for this node. More...

Additional Inherited Members
Public Attributes inherited from ONNXOperatorParser
	node_id

	node

	nodes_outputs

	onnx_model

	mapping_data

	accelerator

Static Public Attributes inherited from ONNXOperatorParser
string	CUSTOM_WEIGHT_SIZE_ATTR = "weight_size"

string	CUSTOM_ACT_SIZE_ATTR = "act_size"

string	CUSTOM_OUTPUT_SIZE_ATTR = "output_size"

Detailed Description

Parses an ONNX Gemm operator into a LayerNode.

Member Function Documentation

◆ generate_layer_node()

def generate_layer_node ( self )

Here is the call graph for this function:

Here is the caller graph for this function:

◆ get_layer_node_user_format()

dict[str, Any] get_layer_node_user_format	(		self,
		list[int]	input_shape,
		list[int]	output_shape
	)

Generate layer data in user input format for MatMul or GEMM ONNX nodes.

I[B] [D][C] * W([B]) [C][K]-> O [B] [D][K] or I[B][H][D][C] * W([B][H])[C][K]-> O [B][H][D][K]

Here is the call graph for this function:

Here is the caller graph for this function:

◆ infer_input_activation_shape()

list[int] infer_input_activation_shape	(		self,
		list[int]	output_shape
	)

In case the input activations are empty (which can happen if there is a shape operator in the path): extract the weights from the model graph initializer to get the correct input activation size.

NOTE: only implemented for tensors with dimension size 2.