xref: /aosp_15_r20/external/tensorflow/tensorflow/python/eager/profiler.py (revision b6fb3261f9314811a0f4371741dbb8839866f948) 
1# Copyright 2017 The TensorFlow Authors. All Rights Reserved.
2#
3# Licensed under the Apache License, Version 2.0 (the "License");
4# you may not use this file except in compliance with the License.
5# You may obtain a copy of the License at
6#
7#     http://www.apache.org/licenses/LICENSE-2.0
8#
9# Unless required by applicable law or agreed to in writing, software
10# distributed under the License is distributed on an "AS IS" BASIS,
11# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12# See the License for the specific language governing permissions and
13# limitations under the License.
14# ==============================================================================
15"""TensorFlow 2.0 Profiler for both Eager Mode and Graph Mode.
16
17The profiler has two mode:
18- Programmatic Mode: start(), stop() and Profiler class. It will perform
19                    when calling start() or create Profiler class and will stop
20                    when calling stop() or destroying Profiler class.
21- On-demand Mode: start_profiler_server(). It will perform profiling when
22                  receive profiling request.
23
24NOTE: Only one active profiler session is allowed. Use of simultaneous
25Programmatic Mode and On-demand Mode is undefined and will likely fail.
26
27NOTE: The Keras TensorBoard callback will automatically perform sampled
28profiling. Before enabling customized profiling, set the callback flag
29"profile_batches=[]" to disable automatic sampled profiling.
30customized profiling.
31"""
32
33import datetime
34import os
35import threading
36
37from tensorflow.python.client import _pywrap_events_writer
38from tensorflow.python.eager import context
39from tensorflow.python.framework import errors
40from tensorflow.python.platform import gfile
41from tensorflow.python.platform import tf_logging as logging
42from tensorflow.python.profiler.internal import _pywrap_profiler
43from tensorflow.python.util import compat
44from tensorflow.python.util.deprecation import deprecated
45
46_profiler = None
47_profiler_lock = threading.Lock()
48_run_num = 0
49# This suffix should be kept in sync with kProfileEmptySuffix in
50# tensorflow/core/profiler/rpc/client/capture_profile.cc.
51_EVENT_FILE_SUFFIX = '.profile-empty'
52
53
54class ProfilerAlreadyRunningError(Exception):
55  pass
56
57
58class ProfilerNotRunningError(Exception):
59  pass
60
61
62@deprecated('2020-07-01', 'use `tf.profiler.experimental.start` instead.')
63def start(options=None):
64  """Start profiling.
65
66  Args:
67    options: profiler options.
68
69  Raises:
70    ProfilerAlreadyRunningError: If another profiling session is running.
71  """
72  global _profiler
73  with _profiler_lock:
74    if _profiler is not None:
75      raise ProfilerAlreadyRunningError('Another profiler is running.')
76    if context.default_execution_mode == context.EAGER_MODE:
77      context.ensure_initialized()
78    _profiler = _pywrap_profiler.ProfilerSession()
79    try:
80      _profiler.start('', options if options is not None else {})
81    except errors.AlreadyExistsError:
82      logging.warning('Another profiler session is running which is probably '
83                      'created by profiler server. Please avoid using profiler '
84                      'server and profiler APIs at the same time.')
85      raise ProfilerAlreadyRunningError('Another profiler is running.')
86
87
88@deprecated('2020-07-01', 'use `tf.profiler.experimental.stop` instead.')
89def stop():
90  """Stop current profiling session and return its result.
91
92  Returns:
93    A binary string of tensorflow.tpu.Trace. User can write the string
94    to file for offline analysis by tensorboard.
95
96  Raises:
97    ProfilerNotRunningError: If there is no active profiling session.
98  """
99  global _profiler
100  global _run_num
101  with _profiler_lock:
102    if _profiler is None:
103      raise ProfilerNotRunningError(
104          'Cannot stop profiling. No profiler is running.')
105    if context.default_execution_mode == context.EAGER_MODE:
106      context.context().executor.wait()
107    result = _profiler.stop()
108    _profiler = None
109    _run_num += 1
110  return result
111
112
113@deprecated(
114    '2020-07-01',
115    '`tf.python.eager.profiler` has deprecated, use `tf.profiler` instead.'
116)
117def maybe_create_event_file(logdir):
118  """Create an empty event file if not already exists.
119
120  This event file indicates that we have a plugins/profile/ directory in the
121  current logdir.
122
123  Args:
124    logdir: log directory.
125  """
126  for file_name in gfile.ListDirectory(logdir):
127    if file_name.endswith(_EVENT_FILE_SUFFIX):
128      return
129  # TODO(b/127330388): Use summary_ops_v2.create_file_writer instead.
130  event_writer = _pywrap_events_writer.EventsWriter(
131      compat.as_bytes(os.path.join(logdir, 'events')))
132  event_writer.InitWithSuffix(compat.as_bytes(_EVENT_FILE_SUFFIX))
133
134
135@deprecated(
136    '2020-07-01',
137    '`tf.python.eager.profiler` has deprecated, use `tf.profiler` instead.'
138)
139def save(logdir, result):
140  """Save profile result to TensorBoard logdir.
141
142  Args:
143    logdir: log directory read by TensorBoard.
144    result: profiling result returned by stop().
145  """
146  plugin_dir = os.path.join(
147      logdir, 'plugins', 'profile',
148      datetime.datetime.now().strftime('%Y-%m-%d_%H-%M-%S'))
149  gfile.MakeDirs(plugin_dir)
150  maybe_create_event_file(logdir)
151  with gfile.Open(os.path.join(plugin_dir, 'local.trace'), 'wb') as f:
152    f.write(result)
153
154
155@deprecated('2020-07-01', 'use `tf.profiler.experimental.server.start`.')
156def start_profiler_server(port):
157  """Start a profiler grpc server that listens to given port.
158
159  The profiler server will keep the program running even the training finishes.
160  Please shutdown the server with CTRL-C. It can be used in both eager mode and
161  graph mode. The service defined in
162  tensorflow/core/profiler/profiler_service.proto. Please use
163  tensorflow/contrib/tpu/profiler/capture_tpu_profile to capture tracable
164  file following https://cloud.google.com/tpu/docs/cloud-tpu-tools#capture_trace
165
166  Args:
167    port: port profiler server listens to.
168  """
169  if context.default_execution_mode == context.EAGER_MODE:
170    context.ensure_initialized()
171  _pywrap_profiler.start_server(port)
172
173
174@deprecated('2020-07-01', 'use `tf.profiler.experimental.Profile` instead.')
175class Profiler(object):
176  """Context-manager eager profiler api.
177
178  Example usage:
179  ```python
180  with Profiler("/path/to/logdir"):
181    # do some work
182  ```
183  """
184
185  def __init__(self, logdir):
186    self._logdir = logdir
187
188  def __enter__(self):
189    start()
190
191  def __exit__(self, typ, value, tb):
192    result = stop()
193    save(self._logdir, result)
194