class PreOptimization():
def __init__(self, model, optimization):
self.model = model
self.optimization = optimization
self.output_node_names = model.output_node_names
self.input_node_names = model.input_node_names
if model.iter_op is not None:
self.output_node_names.append('MakeIterator')
self.analyzer = GraphAnalyzer()
self.analyzer.graph = model.graph_def
self.analyzer.parse_graph()
self.logger = logging.getLogger()
self._tmp_graph_def = None
self._excluded_node_names = []
def get_excluded_node_names(self):
"""Get the excluded node name
Returns:
string list: the excluded ops' name
"""
return self._excluded_node_names
@dump_elapsed_time("Pass Pre Optimization")
def get_optimized_model(self):
"""Executed the non-precision dependant graph optimization.
The input graph will be optimized with following passes:
1. Remove the training nodes like Identity Op.
2. Split the shared nodes like weights node for multi-Conv2d.
3. Fold Constant Nodes as less as possible.
4. Fuse the Mul node into the previous Conv2D/MatMul if possible.
5. Strip the useless nodes.
6. Do the Common sequence elimation optimization on the graph.
7. Fold the BN node into the previous Conv2D if possible.
Returns:
[graphdef]: the optimized graphdef object.
"""
self.logger.debug("Start to pre optimize input model...")
origin_model = TensorflowModel(self.model._model,
self.model.framework_specific_info,
**self.model.kwargs)
self._tmp_graph_def = ConvertLayoutOptimizer(
self.model.graph_def, self.output_node_names).do_transformation()
self._tmp_graph_def = GrapplerOptimizer(
self._tmp_graph_def, self.output_node_names, self.optimization).do_transformation()
self._tmp_graph_def = RemoveTrainingNodesOptimizer(
self._tmp_graph_def, protected_nodes=self.output_node_names).do_transformation()
self._tmp_graph_def = SplitSharedInputOptimizer(self._tmp_graph_def).do_transformation()
self._tmp_graph_def = GraphFoldConstantOptimizer(self._tmp_graph_def).do_transformation()
self._tmp_graph_def = FuseColumnWiseMulOptimizer(self._tmp_graph_def).do_transformation()
self._tmp_graph_def = StripUnusedNodesOptimizer(self._tmp_graph_def,
self.input_node_names, self.output_node_names).do_transformation()
self._tmp_graph_def = FuseGeluOptimizer(self._tmp_graph_def).do_transformation()
self._tmp_graph_def = GraphCseOptimizer(self._tmp_graph_def).do_transformation()
self._tmp_graph_def = FoldBatchNormNodesOptimizer(
self._tmp_graph_def).do_transformation()
#TODO we should handle all control ops elegantly not bypass it.
self._tmp_graph_def, excluded_node_names = UpdateEnterOptimizer(
self._tmp_graph_def).do_transformation()
#TODO we need to remove below optimizer once the TF enabled the single
# matmul op quantization
self._tmp_graph_def = InjectDummyBiasAddOptimizer(
self._tmp_graph_def).do_transformation()
self._tmp_graph_def = FuseTransposeReshapeOptimizer(
self._tmp_graph_def).do_transformation()
self._excluded_node_names.extend(excluded_node_names)
self._tmp_graph_def.library.CopyFrom(self.model.graph_def.library)
origin_model.graph_def = self._tmp_graph_def
return origin_model
def get_matched_nodes(self, patterns):
"""Searche the matched nodes with the specified patterns
Args:
patterns ([string list]): The patterns should be illustrated as below.
[['MatMul'], ("BiasAdd"), ("Relu",)]
Returns:
[string list]: It will return the list that contains the matched nodes name
and pattern. ['matched_node_a_name', 'matched_node_a_name',['MatMul','BiasAdd']]
"""
self.analyzer.graph = self._tmp_graph_def
self.analyzer.parse_graph()
res = []
for sub_pattern in patterns:
res.extend([i for i in self.analyzer.query_fusion_pattern_nodes(
sub_pattern) if i not in res])
return res
def has_positive_input(self, node_name):
"""Check the specified node has the positive input or not.
Args:
node_name ([string]): node's name
Returns:
[bool]: True if the node has the positive input data,
False if the node has the negative input data.
"""
return self.analyzer.has_positive_input(node_name)
class PreOptimization(object):
def __init__(self, model, inputs, outputs):
self.output_node_names = list(
set([output.split(":")[0] for output in outputs]))
self.input_graph = get_graph_def(model, self.output_node_names)
if 'MakeIterator' in [node.op for node in self.input_graph.node]:
self.output_node_names.append('MakeIterator')
self.analyzer = GraphAnalyzer()
self.analyzer.graph = self.input_graph
self.analyzer.parse_graph()
self.input_node_names = inputs
self.logger = logging.getLogger()
self._tmp_graph_def = None
self._excluded_node_names = []
if not self.input_node_names or not self.output_node_names:
self.input_node_names, self.output_node_names = self.analyzer.get_graph_input_output(
)
def get_excluded_node_names(self):
"""Get the excluded node name
Returns:
string list: the excluded ops' name
"""
return self._excluded_node_names
@dump_elapsed_time("Pass Pre Optimization")
def get_optimized_graphdef(self):
"""Executed the non-precision dependant graph optimization.
The input graph will be optimized with following passes:
1. Remove the training nodes like Identity Op.
2. Split the shared nodes like weights node for multi-Conv2d.
3. Fold Constant Nodes as less as possible.
4. Fuse the Mul node into the previous Conv2D/MatMul if possible.
5. Strip the useless nodes.
6. Do the Common sequence elimation optimization on the graph.
7. Fold the BN node into the previous Conv2D if possible.
Returns:
[graphdef]: the optimized graphdef object.
"""
self.logger.debug("Start to pre optimize input model...")
self._tmp_graph_def = ConvertLayoutOptimizer(
self.input_graph, self.output_node_names).do_transformation()
self._tmp_graph_def = RemoveTrainingNodesOptimizer(
self._tmp_graph_def,
protected_nodes=self.output_node_names).do_transformation()
self._tmp_graph_def = SplitSharedInputOptimizer(
self._tmp_graph_def).do_transformation()
self._tmp_graph_def = GraphFoldConstantOptimizer(
self._tmp_graph_def).do_transformation()
self._tmp_graph_def = FuseColumnWiseMulOptimizer(
self._tmp_graph_def).do_transformation()
self._tmp_graph_def = StripUnusedNodesOptimizer(
self._tmp_graph_def, self.input_node_names,
self.output_node_names).do_transformation()
self._tmp_graph_def = GraphCseOptimizer(
self._tmp_graph_def).do_transformation()
self._tmp_graph_def = FoldBatchNormNodesOptimizer(
self._tmp_graph_def).do_transformation()
#TODO we should handle all control ops elegantly not bypass it.
self._tmp_graph_def, excluded_node_names = UpdateEnterOptimizer(
self._tmp_graph_def).do_transformation()
self._excluded_node_names.extend(excluded_node_names)
self._tmp_graph_def.library.CopyFrom(self.input_graph.library)
return self._tmp_graph_def
def get_matched_nodes(self, patterns):
"""Searche the matched nodes with the specified patterns
Args:
patterns ([string list]): The patterns should be illustrated as below.
[['MatMul'], ("BiasAdd"), ("Relu",)]
Returns:
[string list]: It will return the list that contains the matched nodes name
and pattern. ['matched_node_a_name', 'matched_node_a_name',['MatMul','BiasAdd']]
"""
self.analyzer.graph = self._tmp_graph_def
self.analyzer.parse_graph()
res = []
for sub_pattern in patterns:
res.extend(self.analyzer.query_fusion_pattern_nodes(sub_pattern))
return res
def has_positive_input(self, node_name):
"""Check the specified node has the positive input or not.
Args:
node_name ([string]): node's name
Returns:
[bool]: True if the node has the positive input data,
False if the node has the negative input data.
"""
return self.analyzer.has_positive_input(node_name)
def test_graph_cse(self):
tf.compat.v1.disable_eager_execution()
input_constant_name = "input_constant"
relu_name = "relu"
float_graph_def = graph_pb2.GraphDef()
input_constant = QuantizeGraphHelper.create_constant_node(
input_constant_name,
value=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
dtype=dtypes.float32,
shape=[1, 2, 6, 1])
float_graph_def.node.extend([input_constant])
relu_node = QuantizeGraphHelper.create_node("Relu", relu_name,
[input_constant_name])
QuantizeGraphHelper.set_attr_dtype(relu_node, "T", dtypes.float32)
float_graph_def.node.extend([relu_node])
b_constant_name = "b_constant"
mat_mul_name = "mat_mul"
b_constant = QuantizeGraphHelper.create_constant_node(
b_constant_name,
value=[1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12],
dtype=dtypes.float32,
shape=[2, 6])
float_graph_def.node.extend([b_constant])
mat_mul_node = QuantizeGraphHelper.create_node(
"MatMul", mat_mul_name, [relu_name, b_constant_name])
QuantizeGraphHelper.set_attr_dtype(mat_mul_node, "T", dtypes.float32)
QuantizeGraphHelper.set_attr_bool(mat_mul_node, "transpose_a", False)
QuantizeGraphHelper.set_attr_bool(mat_mul_node, "transpose_b", False)
float_graph_def.node.extend([mat_mul_node])
bias_add_name = "bias_add"
offset_constant_name = "offset_constant"
offset_constant = QuantizeGraphHelper.create_constant_node(
offset_constant_name,
value=[1, 2, 3, 4, 5, 6],
dtype=dtypes.float32,
shape=[6])
float_graph_def.node.extend([offset_constant])
bias_add_node = QuantizeGraphHelper.create_node(
"BiasAdd", bias_add_name, [mat_mul_name, offset_constant_name])
QuantizeGraphHelper.set_attr_dtype(bias_add_node, "T", dtypes.float32)
float_graph_def.node.extend([bias_add_node])
post_relu_name = "post_relu"
post_relu_node = QuantizeGraphHelper.create_node(
"Relu", post_relu_name, [bias_add_name])
float_graph_def.node.extend([post_relu_node])
last_identity_node_name = 'last_identity'
last_identity_node = QuantizeGraphHelper.create_node(
"Identity", last_identity_node_name, [post_relu_name])
float_graph_def.node.extend([last_identity_node])
analyzer = GraphAnalyzer()
analyzer.graph = float_graph_def
analyzer.parse_graph()
res = analyzer.query_fusion_pattern_nodes([['MatMul'], ("BiasAdd"),
("Relu")])
self.assertEqual(3, len(res[0][-1]))