def create_metric_reporter(cls, config: Config,
tensorizers: Dict[str, Tensorizer]):
return NERMetricReporter(
channels=[ConsoleChannel()],
label_names=list(tensorizers["tokens"].labels_vocab._vocab),
pad_idx=tensorizers["tokens"].labels_pad_idx,
)
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
return cls(
[
ConsoleChannel(),
Seq2SeqFileChannel([Stage.TEST], config.output_path, tensorizers),
],
tensorizers,
)
def from_config(cls, config, *args, tensorizers=None, **kwargs):
return cls(
channels=[
ConsoleChannel(),
MultiSpanQAFileChannel((Stage.TEST, ), config.output_path),
],
tensorizer=tensorizers["tokens"],
)
def from_config(cls, config, *args, tensorizers=None, **kwargs):
return cls(
channels=[
ConsoleChannel(),
FileChannel((Stage.TEST, ), config.output_path)
],
text_column_names=config.text_column_names,
model_select_metric=config.model_select_metric,
task_batch_size=config.task_batch_size,
num_negative_ctxs=config.num_negative_ctxs,
)
def from_config(cls, config, *args, tensorizers=None, **kwargs):
return cls(
channels=[
ConsoleChannel(),
SquadFileChannel((Stage.TEST, ), config.output_path),
],
n_best_size=config.n_best_size,
max_answer_length=config.max_answer_length,
ignore_impossible=config.ignore_impossible,
has_answer_labels=tensorizers["has_answer"].vocab._vocab,
tensorizer=tensorizers["squad_input"],
false_label=config.false_label,
)
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
channels = [ConsoleChannel()]
if config.TEMP_DUMP_PREDICTIONS:
channels.append(
Seq2SeqFileChannel([Stage.TEST], config.output_path, tensorizers),
)
return cls(
channels,
config.log_gradient,
tensorizers,
config.model_select_metric_key,
config.select_length_beam,
config.print_length_metrics,
)
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
return cls(
[
ConsoleChannel(),
CompositionalSeq2SeqFileChannel(
[Stage.TEST],
config.output_path,
tensorizers,
config.accept_flat_intents_slots,
),
],
tensorizers,
config.accept_flat_intents_slots,
)
def from_config(cls, config: Config, tensorizers: Dict[str, Tensorizer]):
channels = [ConsoleChannel()]
if config.TEMP_DUMP_PREDICTIONS:
channels.append(
MaskedCompositionalSeq2SeqFileChannel(
[Stage.TEST],
config.output_path,
tensorizers,
config.accept_flat_intents_slots,
))
return cls(
channels,
config.log_gradient,
tensorizers,
config.accept_flat_intents_slots,
config.model_select_metric_key,
config.select_length_beam,
)
def from_config(cls, config, tensorizers):
return MyTaggingMetricReporter(
channels=[ConsoleChannel(), TensorBoardChannel()],
label_names=tensorizers["slots"].vocab,
)
def from_config0(cls, config, vocab):
return MyTaggingMetricReporter(
channels=[ConsoleChannel(), TensorBoardChannel()],
label_names=vocab)
def create_metric_reporter(cls, config, tensorizers):
return MyTaggingMetricReporter(
channels=[ConsoleChannel(), TensorBoardChannel()],
label_names=list(tensorizers["slots"].vocab),
)
def from_config(cls, config, *args, **kwargs):
return cls([ConsoleChannel()], config.pep_format)
def from_config(cls, config: PyTextConfig, pad_index: int = -1):
return cls(channels=[ConsoleChannel()], pad_index=pad_index)