Source code for experimental_experiment.xoptim.patterns_ort.gather_grad
import inspect
from typing import List, Optional
from onnx import NodeProto
from onnx.numpy_helper import to_array
from ..patterns_api import MatchResult, PatternOptimization
[docs]
class GatherGradPattern(PatternOptimization):
"""
Replaces ConstantOfShape + ScatterND with GatherGrad (com.domain).
"""
[docs]
def match(
self,
g: "GraphBuilderPatternOptimization", # noqa: F821
node: NodeProto,
matched: List[MatchResult],
) -> Optional[MatchResult]:
if node.op_type != "ScatterND" or node.domain != "":
return self.none()
reduction = g.get_attribute(node, "reduction")
if reduction is None or reduction.s != b"add":
return self.none(node, inspect.currentframe().f_lineno)
if not g.has_type(node.input[2]):
itype = g.try_infer_type(node.input[2])
if itype == 0:
return self.none(node, inspect.currentframe().f_lineno)
else:
itype = g.get_type(node.input[2])
node_before = g.node_before(node.input[0])
if node_before.op_type != "ConstantOfShape" or node.domain != "":
return self.none(node, inspect.currentframe().f_lineno)
att = g.get_attribute(node_before, "value", False)
if att is not None:
arr = to_array(att.t)
if arr[0] != 0:
return self.none(node, inspect.currentframe().f_lineno)
return MatchResult(self, [node_before, node], self.apply, insert_at=node)
[docs]
def apply(
self,
g: "GraphBuilder", # noqa: F821
node_before: NodeProto,
node: NodeProto,
) -> List[NodeProto]:
new_node = g.make_node(
"GatherGrad",
[node_before.input[0], *node.input[1:]],
node.output,
name=f"{self.__class__.__name__}--{node.name}",
domain="com.microsoft",
)
for att in node.attribute:
if att.name != "reduction":
new_node.append(att)
return [new_node]