seqio.loggers package

Classes for logging evaluation metrics and inference results.

class seqio.loggers.JSONLogger(output_dir, write_n_results=None, json_encoder_cls=<class 'seqio.loggers.TensorAndNumpyEncoder'>)

A logger that writes metrics and model outputs to JSONL files.

class seqio.loggers.Logger(output_dir)

Abstract base class for logging.

output_dir

a directory to save the logging results (e.g., TensorBoard summary) as well as the evaluation results (e.g., “inputs_pretokenized”, “target_pretokenize” and “prediction”).

class seqio.loggers.PyLoggingLogger(output_dir, level=0)

A logger that writes metrics using the standard Python log.

class seqio.loggers.TensorAndNumpyEncoder(*args, max_ndarray_size=32, **kwargs)

JSON Encoder to use when encoding dicts with tensors and numpy arrays.

default(obj)

Implement this method in a subclass such that it returns a serializable object for o, or calls the base implementation (to raise a TypeError).

For example, to support arbitrary iterators, you could implement default like this:

def default(self, o):
    try:
        iterable = iter(o)
    except TypeError:
        pass
    else:
        return list(iterable)
    # Let the base class default method raise the TypeError
    return JSONEncoder.default(self, o)

class seqio.loggers.TensorBoardLogger(output_dir)

A logger that writes metrics to TensorBoard summaries.

class seqio.loggers.TensorBoardLoggerV1(output_dir)

A logger that writes metrics to TensorBoard summaries in TF1.

seqio.loggers.skip_none_value_dict_factory(data)

Dictionnary factory which skip None value.