def test_id_match(patrn_ugraph):
matcher = uTensorGraphMatcher(
patrn_ugraph, op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(patrn_ugraph)
assert matches, 'expecting matches, get {} matches'.format(len(matches))
match = matches[0]
assert match.patrn2subj_op_map['input0'].name in ['input0', 'input1']
assert match.patrn2subj_op_map['input1'].name in ['input0', 'input1']
assert match.patrn2subj_op_map['input0'].name != match.patrn2subj_op_map[
'input1'].name
assert match.patrn2subj_op_map['add0'].name == 'add0'
assert match.patrn2subj_op_map['output'].name == 'output'
assert match.subj2patrn_op_map['input0'].name in ['input0', 'input1']
assert match.subj2patrn_op_map['input1'].name in ['input0', 'input1']
assert match.subj2patrn_op_map['input0'].name != match.subj2patrn_op_map[
'input1'].name
assert match.subj2patrn_op_map['add0'].name == 'add0'
assert match.subj2patrn_op_map['output'].name == 'output'
for tensor in patrn_ugraph.input_tensors:
assert tensor.name in match.patrn2subj_tensor_map, \
'{} is missing'.format(tensor.name)
for tensor in patrn_ugraph.output_tensors:
assert tensor.name in match.subj2patrn_tensor_map, \
'{} is missing'.format(tensor.name)
def test_replace_fc_with_add(subj_graph_1, patrn_fc_1):
def callback(match):
graph = tf.Graph()
with graph.as_default():
a = tf.placeholder(dtype=tf.float32, name='a')
b = tf.placeholder(dtype=tf.float32, name='b')
out = tf.add(a, b, name='fused_node')
ugraph = GraphDefParser(config={}).parse(graph.as_graph_def(), output_nodes=[out.op.name])
ugraph.ops_info['fused_node'].replace_with_null_input_tensor(0)
ugraph.ops_info['fused_node'].replace_with_null_input_tensor(1)
topologic_order_graph(ugraph)
ugraph = prune_graph(ugraph)
patrn_ugraph = match.pattern_ugraph
input_map = {
patrn_ugraph.ops_info['a_prime'].input_tensors[0]: ugraph.ops_info['fused_node'].input_tensors[0],
patrn_ugraph.ops_info['a_prime'].input_tensors[1]: ugraph.ops_info['fused_node'].input_tensors[1]
}
output_map = {
patrn_ugraph.ops_info['r_prime'].output_tensors[0]: ugraph.ops_info['fused_node'].output_tensors[0]
}
return ugraph, input_map, output_map
matcher = uTensorGraphMatcher(patrn_fc_1, op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(subj_graph_1)
assert matches, 'no match found'
match = matches[0]
new_ugraph = match.replace_with(callback)
test_pass = True
missed_op_names = []
for op_name in match.subj2patrn_op_map:
if op_name in new_ugraph.ops_info:
test_pass = False
missed_op_names.append(op_name)
assert test_pass, \
'these ops should not be found in the new ugrah: {}'.format(missed_op_names)
def transform(self, ugraph):
# FIXME: should use a generic op_equality_delegate
matcher = uTensorGraphMatcher(
pattern_ugraph=self.pattern_ugraph,
op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(ugraph, 1)
while matches:
match = matches[0]
ugraph = match.replace_with(callback=self)
matches = matcher.match(ugraph, 1)
return ugraph
def _transform_tf(self, ugraph):
# FIXME: should use a generic op_equality_delegate
matcher = uTensorGraphMatcher(
pattern_ugraph=self.pattern_ugraph,
op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(ugraph, n=1)
while matches:
match = matches[0]
ugraph = self._handle_match_tf(match)
matches = matcher.match(ugraph)
return ugraph
def transform(self, ugraph):
if ugraph.lib_name != 'tensorflow':
raise ValueError('only support tensorflow graph')
# FIXME: should use a generic op_equality_delegate
matcher = uTensorGraphMatcher(
pattern_ugraph=self.pattern_ugraph,
op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(ugraph, n=1)
while matches:
match = matches[0]
ugraph = match.replace_with(callback=self)
matches = matcher.match(ugraph, n=1)
return ugraph
def test_match_sub1_2(patrn_ugraph, subject_ugraph1_2):
matcher = uTensorGraphMatcher(
patrn_ugraph, op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match(subject_ugraph1_2)
assert matches, 'expecting matches, get {} matches'.format(len(matches))
match = matches[0]
assert match.patrn2subj_op_map['input0'].name in [
'sub_input0', 'sub_input1'
]
assert match.patrn2subj_op_map['input1'].name in [
'sub_input0', 'sub_input1'
]
assert match.patrn2subj_op_map['add0'].name == 'sub_add0'
assert match.patrn2subj_op_map['output'].name == 'sub_add1'
def test_match_sub1(patrn_ugraph, subject_ugraph1):
matcher = uTensorGraphMatcher(
patrn_ugraph, op_equality_delegate=uTensorOpEqualityDelegate)
matches = matcher.match_all(subject_ugraph1)
assert matches, 'expecting matches, get {} matches'.format(len(matches))
match = matches[0]
assert len(matches) == 2, 'should be exactly two match, get {}'.format(
len(matches))
assert match.patrn2subj_op_map['input0'].name in [
'sub_input0', 'sub_input1'
], match
assert match.patrn2subj_op_map['input1'].name in [
'sub_input0', 'sub_input1'
], match
assert match.patrn2subj_op_map['input0'].name != match.patrn2subj_op_map[
'input1'].name
assert match.patrn2subj_op_map['add0'].name == 'sub_add0', match
assert match.patrn2subj_op_map['output'].name == 'sub_add1', match