1.7.1

What's Changed

• Added apply_timed_decorator for applying timing decorators to arbitrary functions without changing source code.
• Supports default timed decorator and registered named decorators via create_timed_decorator.
• Named decorators are lazily resolved, consistent with get_timed_decorator.
• Updated tests.
• Fixed new flake8 errors.

Full Changelog: 1.6.1...1.7.1

1.6.1

What's Changed

• Returning seconds instead of nanoseconds when return_time=True and use_seconds=True.

Full Changelog: 1.6.0...1.6.1

1.6.0

What's Changed

• Timed decorator can now operate with function objects.
• Time decorator now works even if the __name__ or __qualname__ attribute is missing for the timed Callable.
• Fixed test for timing cuda tensors

Full Changelog: 1.5.2...1.6.0

1.5.2

What's Changed

• Timed decorator can now operate with torch.jit.ScriptFunctions in #7.

Full Changelog: 1.5.1...1.5.2

1.5.1

What's Changed

• Added parameter to disable created decorator in #6.

Full Changelog: 1.5.0...1.5.1

1.5.0

What's Changed

• Changing behavior of out parameter by in #5.
  • Previous behavior: if out is passed to the decorator, the elapsed time is stored in the dict using the fully qualified name as key.
  • New behavior: if out is passed to the decorator, the number of function calls, the elapsed time and the "own time" is stored using the fully qualified name as key.

Full Changelog: 1.4.0...1.5.0

1.4.0

What's Changed

New parameter for using the fully qualified name of the function

• The fully qualified name of the function (__qualname__) is used for logging when use_qualname is enabled.
• The out parameter now uses the function's fully qualified name as key.

Added new facility for registering timed decorators and using the same decorator throughout the codebase

• This allows creating a complicated timer which is also able to accumulate elapsed times from all the measured functions. It can be easily manipulated to create code statistics.

Full Changelog: 1.3.0...1.4.0

1.3.0

What's Changed

Diversified measuring and printing options.

• Changed behavior when printing elapsed time
  • Previous behavior: elapsed time is either printed to stdout, written to file or to logger.
  • Current behavior: printing, logging and writing are not mutually exclusive anymore. Each is controlled by its own parameter.
• Added new parameter for returning elapsed time in addition to the function's return value.
• Updated classifiers in pyproject.toml.
• Improved test speed.

Full Changelog: 1.2.2...1.3.0

1.2.2

What's Changed

• Added option to write execution time to a dict.
• Added option to disable garbage collection during function execution.

Full Changelog: 1.2.0...1.2.2

1.2.0

Added new logging options.
Measurements can be logged to a file or to a logger using the python logging facilities.

Full Changelog: 1.1.1...1.2.0