Add initial code

Add mllog package, adapted from mlperf/training

.gitignore

@@ -0,0 +1,129 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+pip-wheel-metadata/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+.python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/

README.md

@@ -0,0 +1 @@
+# MLPerf Compliance Logging Utilities and Helper Functions

mllog/README.md

@@ -0,0 +1,31 @@
+# MLLog
+
+## Install
+
+Use one of the following ways to install.
+
+- For development, you may download the latest version and install from local path:
+
+```sh
+git clone https://github.com/mlperf/logging.git
+pip install -e logging
+```
+
+- Install from github at a specific commit (Replace COMMIT_HASH with actual commit hash, the double quotes are needed):
+  ```
+  pip install "git+https://github.com/mlperf/logging@COMMIT_HASH"
+  ```
+
+Uninstall:
+
+```sh
+pip uninstall mlperf_compliance
+```
+
+## Use
+
+[Here](examples/dummy_example.py) is an example of how to use the `mllog` package in benchmark codes. You can run the example by:
+
+```sh
+python3 -m mllog.examples.dummy_example 
+```

mllog/__init__.py

@@ -0,0 +1,97 @@
+# Copyright 2019 MLBenchmark Group. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+
+import logging
+import sys
+import threading
+
+from mllog import mllog
+
+
+# _lock is for serializing access to shared data structures in this module.
+_lock = threading.RLock()
+
+# mllogger is a logger shareable across modules.
+mllogger = mllog.MLLogger()
+
+
+def get_mllogger():
+  """Get the shared logger."""
+  return mllogger
+
+def config(**kwargs):
+  """Configure the shared logger.
+  Optional keyword arguments:
+    logger: a logging.Logger instance. Customize the logger to change
+      the logging behavior (e.g. logging to a file, etc.)
+    filename: a log file to use. If set, a default file handler will be added
+      to the logger so it can log to the specified file. For more advanced
+      customizations, please set the 'logger' parameter instead.
+    default_namespace: the default namespace to use if one isn't provided.
+    default_stack_offset: the default depth to go into the stack to find the
+      call site.
+    default_clear_line: the default behavior of line clearing (i.e. print
+      an extra new line to clear any pre-existing text in the log line).
+    root_dir: directory prefix which will be trimmed when reporting calling
+      file for logging.
+  """
+  if _lock:
+    _lock.acquire()
+  try:
+    logger = kwargs.pop("logger", None)
+    if logger is not None:
+      if not isinstance(logger, logging.Logger):
+        raise ValueError("'logger' must be an instance of 'logging.Logger'.")
+      if logger.name == mllogger.logger.name:
+        raise ValueError("'logger' should not be the same as the default " +
+                         "logger to avoid unexpected behavior. Consider " +
+                         "using a different name for the logger.")
+      mllogger.logger = logger
+
+    log_file = kwargs.pop("filename", None)
+    if log_file is not None:
+      if not isinstance(log_file, str):
+        raise ValueError("'filename' must be a string.")
+      _file_handler = logging.FileHandler(log_file)
+      _file_handler.setLevel(logging.INFO)
+      mllogger.logger.addHandler(_file_handler)
+
+    default_namespace = kwargs.pop("default_namespace", None)
+    if default_namespace is not None:
+      if not isinstance(default_namespace, str):
+        raise ValueError("'default_namespace' must be a string.")
+      mllogger.default_namespace = default_namespace
+
+    default_stack_offset = kwargs.pop("default_stack_offset", None)
+    if default_stack_offset is not None:
+      if not isinstance(default_stack_offset, int):
+        raise ValueError("'default_stack_offset' must be an integer.")
+      mllogger.default_stack_offset = default_stack_offset
+
+    default_clear_line = kwargs.pop("default_clear_line", None)
+    if default_clear_line is not None:
+      if not isinstance(default_clear_line, bool):
+        raise ValueError("'default_clear_line' must be a boolean value.")
+      mllogger.default_clear_line = default_clear_line
+
+    root_dir = kwargs.pop("root_dir", None)
+    if root_dir is not None:
+      if not isinstance(root_dir, str):
+        raise ValueError("'root_dir' must be a string.")
+      mllogger.root_dir = root_dir
+
+  finally:
+    if _lock:
+      _lock.release()

mllog/constants.py

@@ -0,0 +1,108 @@
+# Copyright 2019 MLBenchmark Group. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================
+
+"""Master list of constants in MLPerf log
+"""
+
+# NOTE: Keep string values in alphabetical order under each section.
+
+# Constant values - log settings
+DEFAULT_LOGGER_NAME = "mllog_default"
+DEFAULT_NAMESPACE = ""
+
+# Constant values - log event type
+INTERVAL_END = "INTERVAL_END"
+INTERVAL_START = "INTERVAL_START"
+POINT_IN_TIME = "POINT_IN_TIME"
+
+# Constant values - benchmark name
+GNMT = "gnmt"
+MASKRCNN = "maskrcnn"
+MINIGO = "minigo"
+NCF = "ncf"
+RESNET = "resnet"
+SSD = "ssd"
+TRANSFORMER = "transformer"
+
+# Constant values - model info
+ADAM = "adam"
+LARS = "lars"
+LAZY_ADAM = "lazy_adam"
+SGD = "sgd"
+
+# Constant values - metadata info
+ABORTED = "aborted"
+SUCCESS = "success"
+
+# Log keys - submission info
+SUBMISSION_BENCHMARK = "submission_benchmark"
+SUBMISSION_DIVISION = "submission_division"
+SUBMISSION_ENTRY = "submission_entry"
+SUBMISSION_ORG = "submission_org"
+SUBMISSION_PLATFORM = "submission_platform"
+SUBMISSION_POC_NAME = "submission_poc_name"
+SUBMISSION_POC_EMAIL = "submission_poc_email"
+SUBMISSION_STATUS = "submission_status"
+
+# Log keys - timing info
+BLOCK_START = "block_start"
+BLOCK_STOP = "block_stop"
+EPOCH_START = "epoch_start"
+EPOCH_STOP = "epoch_stop"
+EVAL_START = "eval_start"
+EVAL_STOP = "eval_stop"
+INIT_START = "init_start"
+INIT_STOP = "init_stop"
+RUN_START = "run_start"
+RUN_STOP = "run_stop"
+
+# Log keys - common run info
+CACHE_CLEAR = "cache_clear"
+EVAL_ACCURACY = "eval_accuracy"
+
+# Log kyes - model hyperparameters
+GLOBAL_BATCH_SIZE = "global_batch_size"
+LARS_EPSILON = "lars_epsilon"
+LARS_OPT_END_LR = "lars_opt_end_learning_rate"
+LARS_OPT_LR_DECAY_POLY_POWER = "lars_opt_learning_rate_decay_poly_power"
+LARS_OPT_LR_DECAY_STEPS = "lars_opt_learning_rate_decay_steps"
+LARS_OPT_WEIGHT_DECAY = "lars_opt_weight_decay"
+MAX_SAMPLES = "max_samples"
+MAX_SEQUENCE_LENGTH = "max_sequence_length"
+MODEL_BN_SPAN = "model_bn_span"
+NUM_IMAGE_CANDIDATES = "num_image_candidates"
+OPT_ADAM_BETA_1 = "opt_adam_beta_1"
+OPT_ADAM_BETA_2 = "opt_adam_beta_2"
+OPT_ADAM_EPSILON = "opt_adam_epsilon"
+OPT_NAME = "opt_name"
+OPT_BASE_LR = "opt_base_learning_rate"
+OPT_LR_ALT_DECAY_FUNC = "opt_learning_rate_alt_decay_func"
+OPT_LR_ALT_WARMUP_FUNC = "opt_learning_rate_alt_warmup_func"
+OPT_LR_DECAY_BOUNDARY_EPOCHS = "opt_learning_rate_decay_boundary_epochs"
+OPT_LR_DECAY_BOUNDARY_STEPS = "opt_learning_rate_decay_boundary_steps"
+OPT_LR_DECAY_FACTOR = "opt_learning_rate_decay_factor"
+OPT_LR_DECAY_INTERVAL = "opt_learning_rate_decay_interval"
+OPT_LR_DECAY_STEPS = "opt_learning_rate_decay_steps"
+OPT_LR_REMAIN_STEPS = "opt_learning_rate_remain_steps"
+OPT_LR_WARMUP_EPOCHS = "opt_learning_rate_warmup_epochs"
+OPT_LR_WARMUP_FACTOR = "opt_learning_rate_warmup_factor"
+OPT_LR_WARMUP_STEPS = "opt_learning_rate_warmup_steps"
+OPT_WEIGHT_DECAY = "opt_weight_decay"
+
+# Log metadata keys
+EPOCH_COUNT = "epoch_count"
+EPOCH_NUM = "epoch_num"
+FIRST_EPOCH_NUM = "first_epoch_num"
+STATUS = "status"

mllog/examples/__init__.py

@@ -0,0 +1,14 @@
+# Copyright 2019 MLBenchmark Group. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+# ==============================================================================