kaolin.utils.log

API

kaolin.utils.log.add_log_level_flag(parser)

Add a log_level flag to an argparser.

Parameters

parser (ArgumentParser) – The argparser to add the flag to.

kaolin.utils.log.default_log_setup(level=20, force=True)

Set up default logging to stdout and quiet optional dependencies.

Parameters

  • level (int) – Logging level for the root logger (default: logging.INFO, i.e. 20).

  • force (bool) – if True (default), will replace any existing loggers with simple stdout.

kaolin.utils.log.log_tensor(t, name, use_logger=None, level=10, print_stats=False, detailed=False)

Log diagnostic tensor information (shape, dtype, optional stats) via a logger.

Uses tensor_info() to format the message.

Parameters

  • t (torch.Tensor or numpy.ndarray or None) – The tensor to describe.

  • name (str) – Human-readable name for the tensor in the log message.

  • use_logger (logging.Logger, optional) – Logger to use. Default: the module logger.

  • level (int) – Logging level (default: logging.DEBUG, i.e. 10).

  • print_stats (bool) – If True, include min/max/mean in the message (default: False).

  • detailed (bool) – If True, include extra tensor properties (default: False).

Examples

>>> t = torch.tensor([1., 2., 3.])
>>> log_tensor(t, 'my_tensor', level=logging.INFO)
kaolin.utils.log.print_tensor(t, name, print_stats=False, detailed=False)

Print diagnostic tensor information (shape, dtype, optional stats) to stdout.

Uses tensor_info() to format the message.

Parameters

  • t (torch.Tensor or numpy.ndarray or None) – The tensor to describe.

  • name (str) – Human-readable name for the tensor in the output.

  • print_stats (bool) – If True, include min/max/mean (default: False).

  • detailed (bool) – If True, include extra tensor properties (default: False).

Examples

>>> t = torch.tensor([1., 2., 3.])
>>> print_tensor(t, 'my_tensor')
my_tensor: torch.Size([3]) (torch.float32)