def profile(graph,
run_meta=None,
op_log=None,
cmd='scope',
options=_DEFAULT_PROFILE_OPTIONS):
"""Print model statistics.
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/profiler/README.md
Args:
graph: tf.Graph.
run_meta: tensorflow::RunMetadata proto. When provided, also shows valid
timing and memory information when 'select' option contains
'micros' and 'bytes'.
op_log: tensorflow::tfprof::OpLog proto. users can use this proto to
group together ops and use a op_type to select the group.
cmd: string. Either 'op', 'scope', 'graph', 'code'.
'op' view organize outputs using operation type. (e.g. MatMul)
'scope' view organize outputs using graph node name scope.
'graph' view organize outputs using graph node inputs/outputs.
'code' view organize outputs using Python call stack.
options: A dict of options. See core/profiler/g3doc/options.md.
Returns:
If cmd is 'scope' or 'graph', returns TFGraphNodeProto proto.
If cmd is 'op' or 'code', returns TFMultiGraphNodeProto proto.
Side effect: stdout/file/timeline.json depending on options['output']
"""
if options == _DEFAULT_PROFILE_OPTIONS:
options = TRAINABLE_VARS_PARAMS_STAT_OPTIONS.copy()
# pylint: disable=protected-access
op_log = tfprof_logger._merge_default_with_oplog(graph,
op_log,
run_meta,
add_trace=cmd == 'code')
# pylint: enable=protected-access
opts = _build_options(options)
run_meta_str = run_meta.SerializeToString() if run_meta else b''
if cmd == 'code' or cmd == 'op':
tfprof_node = tfprof_output_pb2.TFMultiGraphNodeProto()
tfprof_node.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str, op_log.SerializeToString(), cmd.encode('utf-8'),
opts.SerializeToString()))
elif cmd == 'graph' or cmd == 'scope':
tfprof_node = tfprof_output_pb2.TFGraphNodeProto()
tfprof_node.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str, op_log.SerializeToString(), cmd.encode('utf-8'),
opts.SerializeToString()))
else:
raise errors.InvalidArgumentError(None, None,
'unknown cmd: %s\n' % cmd)
return tfprof_node
def advise(graph, run_meta=None, options=_DEFAULT_ADVISE_OPTIONS):
"""Auto profile and advise.
Builds profiles and automatically check anomalies of various
aspects. For more details:
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/profiler/README.md
Args:
graph: required tf.Graph.
run_meta: optional tensorflow.RunMetadata proto. It is necessary to
to support run time information profiling, such as time and memory.
options: see ALL_ADVICE example above. Default checks everything.
Returns:
Returns AdviceProto proto
"""
if options == _DEFAULT_ADVISE_OPTIONS:
options = ALL_ADVICE.copy()
# pylint: disable=protected-access
op_log = tfprof_logger._merge_default_with_oplog(graph,
None,
run_meta,
add_trace=True)
# pylint: enable=protected-access
run_meta_str = run_meta.SerializeToString() if run_meta else b''
opts = _build_advisor_options(options)
ret = tfprof_output_pb2.AdviceProto()
ret.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str, op_log.SerializeToString(), 'advise'.encode('utf-8'),
opts.SerializeToString()))
return ret
def advise(graph, run_meta=None, tfprof_options=ALL_ADVICE): # pylint: disable=dangerous-default-value
"""Auto profile and advise.
Builds profiles and automatically check anormalies of various
aspects. See go/tfprof or README for examples and tutorials.
Args:
graph: tf.Graph.
run_meta: tensorflow::RunMetadata proto. Allows auto-profile
time and memroy.
tfprof_options: see ALL_ADVICE example above.
Returns:
Returns AdviceProto proto
"""
# pylint: disable=protected-access
op_log = tfprof_logger._merge_default_with_oplog(
graph, None, run_meta, add_trace=True)
# pylint: enable=protected-access
run_meta_str = run_meta.SerializeToString() if run_meta else b''
opts = _build_advisor_options(tfprof_options)
ret = tfprof_output_pb2.AdviceProto()
ret.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str,
op_log.SerializeToString(),
'advise'.encode('utf-8'),
opts.SerializeToString()))
return ret
def profile(graph=None,
run_meta=None,
op_log=None,
cmd='scope',
options=_DEFAULT_PROFILE_OPTIONS):
"""Profile model.
Tutorials and examples can be found in:
https://github.com/tensorflow/tensorflow/tree/master/tensorflow/core/profiler/README.md
Args:
graph: tf.Graph. If None and eager execution is not enabled, use
default graph.
run_meta: optional tensorflow.RunMetadata proto. It is necessary to
to support run time information profiling, such as time and memory.
op_log: tensorflow.tfprof.OpLogProto proto. User can assign "types" to
graph nodes with op_log. "types" allow user to flexibly group and
account profiles using options['accounted_type_regexes'].
cmd: string. Either 'op', 'scope', 'graph' or 'code'.
'op' view organizes profile using operation type. (e.g. MatMul)
'scope' view organizes profile using graph node name scope.
'graph' view organizes profile using graph node inputs/outputs.
'code' view organizes profile using Python call stack.
options: A dict of options. See core/profiler/g3doc/options.md.
Returns:
If cmd is 'scope' or 'graph', returns GraphNodeProto proto.
If cmd is 'op' or 'code', returns MultiGraphNodeProto proto.
Side effect: stdout/file/timeline.json depending on options['output']
"""
if not graph and not context.executing_eagerly():
graph = ops.get_default_graph()
if options == _DEFAULT_PROFILE_OPTIONS:
options = (option_builder.ProfileOptionBuilder.
trainable_variables_parameter())
# pylint: disable=protected-access
op_log = tfprof_logger.merge_default_with_oplog(graph,
op_log,
run_meta,
add_trace=cmd == 'code')
# pylint: enable=protected-access
opts = _build_options(options)
run_meta_str = run_meta.SerializeToString() if run_meta else b''
graph_str = _graph_string(graph)
if cmd == 'code' or cmd == 'op':
tfprof_node = tfprof_output_pb2.MultiGraphNodeProto()
ret = print_mdl.PrintModelAnalysis(graph_str, run_meta_str,
op_log.SerializeToString(),
cmd.encode('utf-8'),
opts.SerializeToString())
try:
tfprof_node.ParseFromString(ret)
except message.DecodeError as e:
sys.stderr.write('Cannot parse returned proto: %s.\n' % e)
elif cmd == 'graph' or cmd == 'scope':
tfprof_node = tfprof_output_pb2.GraphNodeProto()
ret = print_mdl.PrintModelAnalysis(graph_str, run_meta_str,
op_log.SerializeToString(),
cmd.encode('utf-8'),
opts.SerializeToString())
try:
tfprof_node.ParseFromString(ret)
except message.DecodeError as e:
sys.stderr.write('Cannot parse returned proto: %s.\n' % e)
else:
raise errors.InvalidArgumentError(None, None,
'unknown cmd: %s\n' % cmd)
return tfprof_node
def testPrintModelAnalysis(self):
opts = tfprof_options_pb2.OptionsProto()
opts.max_depth = TEST_OPTIONS['max_depth']
opts.min_bytes = TEST_OPTIONS['min_bytes']
opts.min_micros = TEST_OPTIONS['min_micros']
opts.min_params = TEST_OPTIONS['min_params']
opts.min_float_ops = TEST_OPTIONS['min_float_ops']
opts.order_by = TEST_OPTIONS['order_by']
opts.step = -1
for p in TEST_OPTIONS['account_type_regexes']:
opts.account_type_regexes.append(p)
for p in TEST_OPTIONS['start_name_regexes']:
opts.start_name_regexes.append(p)
for p in TEST_OPTIONS['trim_name_regexes']:
opts.trim_name_regexes.append(p)
for p in TEST_OPTIONS['show_name_regexes']:
opts.show_name_regexes.append(p)
for p in TEST_OPTIONS['hide_name_regexes']:
opts.hide_name_regexes.append(p)
opts.account_displayed_op_only = TEST_OPTIONS[
'account_displayed_op_only']
for p in TEST_OPTIONS['select']:
opts.select.append(p)
opts.output = TEST_OPTIONS['output']
with session.Session() as sess, ops.device('/cpu:0'):
_ = self._BuildSmallModel()
tfprof_pb = tfprof_output_pb2.TFGraphNodeProto()
tfprof_pb.ParseFromString(
print_mdl.PrintModelAnalysis(
sess.graph.as_graph_def(
add_shapes=True).SerializeToString(), b'', b'',
b'scope', opts.SerializeToString()))
expected_pb = tfprof_output_pb2.TFGraphNodeProto()
text_format.Merge(
r"""name: "_TFProfRoot"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 648
children {
name: "Conv2D"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 2
}
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
}
}
input_shapes {
key: 1
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW"
exec_micros: 0
requested_bytes: 0
parameters: 648
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 648
children {
name: "DW/Assign"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
input_shapes {
key: 1
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW/Initializer"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
children {
name: "DW/Initializer/random_normal"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
children {
name: "DW/Initializer/random_normal/RandomStandardNormal"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 4
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW/Initializer/random_normal/mean"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW/Initializer/random_normal/mul"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
input_shapes {
key: 1
value {
dim {
size: 1
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW/Initializer/random_normal/shape"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
children {
name: "DW/Initializer/random_normal/stddev"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
input_shapes {
key: 1
value {
dim {
size: 1
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 6
}
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 7
}
children {
name: "DW/read"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
input_shapes {
key: 0
value {
dim {
size: 6
}
dim {
size: 6
}
dim {
size: 3
}
dim {
size: 6
}
}
}
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 10
}
children {
name: "zeros"
exec_micros: 0
requested_bytes: 0
total_exec_micros: 0
total_requested_bytes: 0
total_parameters: 0
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 1
}
float_ops: 0
total_float_ops: 0
accelerator_exec_micros: 0
cpu_exec_micros: 0
total_accelerator_exec_micros: 0
total_cpu_exec_micros: 0
run_count: 0
total_run_count: 0
total_definition_count: 13""", expected_pb)
self.assertEqual(expected_pb, tfprof_pb)
def print_model_analysis(graph,
run_meta=None,
op_log=None,
tfprof_cmd='scope',
tfprof_options=TRAINABLE_VARS_PARAMS_STAT_OPTIONS):
"""Print model statistics.
See go/tfprof or README for examples and tutorials.
Run tfprof tool for help:
'bazel run third_party/tensorflow/tools/tfprof help'
Args:
graph: tf.Graph.
run_meta: tensorflow::RunMetadata proto. When provided, also shows valid
timing and memory information when 'select' option contains
'micros' and 'bytes'.
op_log: tensorflow::tfprof::OpLog proto. users can use this proto to
group together ops and use a op_type to select the group.
tfprof_cmd: string. Either 'op', 'scope', 'graph', 'code'.
'op' view organize outputs using operation type. (e.g. MatMul)
'scope' view organize outputs using graph node name scope.
'graph' view organize outputs using graph node inputs/outputs.
'code' view organize outputs using Python call stack.
tfprof_options: See 'tfprof help' for details.
Returns:
If tfprof_cmd is 'scope' or 'graph', returns TFGraphNodeProto proto.
If tfprof_cmd is 'op' or 'code', returns TFMultiGraphNodeProto proto.
Side effect: stdout/file/timeline.json depending on tfprof_options['output']
"""
# pylint: disable=protected-access
op_log = tfprof_logger._merge_default_with_oplog(
graph, op_log, run_meta, add_trace=tfprof_cmd == 'code')
# pylint: enable=protected-access
opts = _build_options(tfprof_options)
run_meta_str = run_meta.SerializeToString() if run_meta else b''
if tfprof_cmd == 'code' or tfprof_cmd == 'op':
tfprof_node = tfprof_output_pb2.TFMultiGraphNodeProto()
tfprof_node.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str,
op_log.SerializeToString(),
tfprof_cmd.encode('utf-8'),
opts.SerializeToString()))
elif tfprof_cmd == 'graph' or tfprof_cmd == 'scope':
tfprof_node = tfprof_output_pb2.TFGraphNodeProto()
tfprof_node.ParseFromString(
print_mdl.PrintModelAnalysis(
graph.as_graph_def(add_shapes=True).SerializeToString(),
run_meta_str,
op_log.SerializeToString(),
tfprof_cmd.encode('utf-8'),
opts.SerializeToString()))
else:
raise errors.InvalidArgumentError(
None, None, 'unknown tfprof_cmd: %s\n' % tfprof_cmd)
return tfprof_node
