compliance_checker

MLPerf training logging compliance checker

Requirements

The checker works with both python2 and python3, requires PyYaml package. See exact versions tested

Usage

To check a log file for compliance:

python -m mlperf_logging.compliance_checker [--config YAML] [--ruleset MLPERF_EDITION] FILENAME

By default, 1.0.0 edition rules are used and the default config is set to 1.0.0/common.yaml. This config will check all common keys and enqueue benchmark specific config to be checked as well. Old editions, still supported are 0.7.0 amd 0.6.0

Prints SUCCESS when no issues were found. Otherwise will print error details.

As log examples use NVIDIA's v0.6 training logs.

Existing config files

1.0.0/common.yaml        - currently the default config file, checks common fields complience and equeues benchmark-specific config file
1.0.0/resnet.yaml
1.0.0/ssd.yaml
1.0.0/minigo.yaml
1.0.0/maskrcnn.yaml
1.0.0/rnnt.yaml
1.0.0/unet3d.yaml
1.0.0/bert.yaml
1.0.0/dlrm.yaml

Implementation details

Compliance checking is done following below algorithm.

1. Parser converts the log into a list of records, each record corresponds to MLLOG line and contains all relevant extracted information
2. Set of rules to be checked in loaded from provided config yaml file
3. Process optional BEGIN rule if present by executing provided CODE section
4. Remove messages for rules that are overridden
5. Loop through the records of the log
   1. If the key in the record is defined in rules process the rule:
      1. If present, execute PRE section
      2. If present, evaluate CHECK section, and store a warning message if the result is false
      3. If present, execute POST section
   2. Increment occurrences counter
6. Store a warning message if any occurrences requirements were violated
7. Process optional END rule if present:
   1. If present, execute PRE
   2. If present, evaluate CHECK section, and raise an exception if the result is false
8. Print all warning messages

Possible side effects of yaml sections execution can be printing output, or enqueueing additional yaml files to be verified.

Config file syntax

Rules to be checked are provided in yaml (config) file. A config file contains the following records:

BEGIN record

Defines CODE to be executed before any other rules defined in the current file. This record is optional and there can be up to a single BEGIN record per config file.

Example:

- BEGIN:
    CODE: " s.update({'run_start':None}) "

KEY record

Defines the actions to be triggered while processing a specific KEY. The name of the KEY is specified in field NAME.

The following fields are optional:

• REQ - specifies the requirement regarding occurrence. Possible values :
  • EXACTLY_ONE - current key has to appear exactly once
  • AT_LEAST_ONE - current key has to appear at least once
  • AT_LEAST_ONE_OR(alternatives) - current key or one of the alternative has to appear at least once; alternatives is a comma separated list of keys
• PRE - code to be executed before performing checks
• CHECK - expression to be evaluated as part of checking this key. False result would mean a failure.
• POST - code to be executed after performing checks

Example:

- KEY:
    NAME:  epoch_start
    REQ:   AT_LEAST_ONE
    CHECK: " s['run_started'] and not s['in_epoch'] and ( v['metadata']['epoch_num'] == (s['last_epoch']+1) ) and not s['run_stopped']"
    POST:  " s['in_epoch'] = True; s['last_epoch'] = v['metadata']['epoch_num'] "

END record

Specifies actions to be taken after processing all the lines in log file. This record is optional and there can be up to a single END record per config file.

The following fields are optional:

• PRE - code to be executed before performing checks
• CHECK - expression to be evaluated as part of checking this key. False result would mean a failure.

Global and local state access

During processing of the records there is a global state s maintained, accessible from code provided in yaml. In addition, rules can access the information fields (values) v of the record, as well as timestamp and the original line string as part of the record ll.

Global state s can be used to enforce any cross keys rules, by updating the global state in POST (or PRE) of one KEY and using that information for CHECK of another KEY. For each config file, s starts as an empty dictionary, so in order to track global state it would require adding an entry to s.

Example:

- BEGIN:
    CODE: " s.update({'run_start':None}) "

ll is a structure representing current log line that triggered KEY record. ll has the following fields that can be accessed:

• full_string - the complete line as a string
• timestamp - milliseconds as an integer
• key - the string key
• value - the parsed value associated with the key, or None if no value
• lineno - line number in the original file of the current key

v is a shortcut for ll.value

Example:

- KEY:
    NAME:  run_stop
    CHECK: " ( v['metadata']['status'] == 'success' )"
    POST:  " print('score [sec]:' , ll.timestamp - s['run_start']) "

Enqueuing additional config files

To enqueue additional rule config files to be verified use enqueue_config(YAML) function. Config files in the queue are processed independently, meaning that they do not share state or any rules.

Each config file may define it's BEGIN and END records, as well as any other KEY rules.

Example:

- KEY:
    NAME:  submission_benchmark
    REQ:   EXACTLY_ONE
    CHECK: " v['value'] in ['resnet', 'ssd', 'maskrcnn', 'transformer', 'gnmt'] "
    POST:  " enqueue_config('1.0.0/{}.yaml'.format(v['value'])) "

Other operations

CODE, REQ, and POST fields are executed using python's exec function. CHECK is performed using eval call. As such, any legal python code would be suitable for use.

For instance, can define rules that would print out information as shown in the example above.

Tested software versions

Tested and confirmed working using the following software versions:

• Python 2.7.12 + PyYAML 3.11
• Python 3.6.8 + PyYAML 5.1
• Python 2.9.2 + PyYAML 5.3.1

How to install PyYaML

pip install pyyaml