def test_registry_has_builtin_similarity_functions(self):
assert SimilarityFunction.by_name(
"dot_product").__name__ == "DotProductSimilarity"
assert SimilarityFunction.by_name(
"bilinear").__name__ == "BilinearSimilarity"
assert SimilarityFunction.by_name(
"linear").__name__ == "LinearSimilarity"
assert SimilarityFunction.by_name(
"cosine").__name__ == "CosineSimilarity"
def from_params(cls, vocab, params: Params) -> 'NlvrCoverageSemanticParser':
sentence_embedder_params = params.pop("sentence_embedder")
sentence_embedder = TextFieldEmbedder.from_params(vocab, sentence_embedder_params)
action_embedding_dim = params.pop_int('action_embedding_dim')
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
dropout = params.pop_float('dropout', 0.0)
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(attention_function_type)
else:
attention_function = None
beam_size = params.pop_int('beam_size')
max_num_finished_states = params.pop_int('max_num_finished_states', None)
normalize_beam_score_by_length = params.pop_bool('normalize_beam_score_by_length', False)
max_decoding_steps = params.pop_int("max_decoding_steps")
checklist_cost_weight = params.pop_float("checklist_cost_weight", 0.6)
dynamic_cost_weight = params.pop("dynamic_cost_weight", None)
penalize_non_agenda_actions = params.pop_bool("penalize_non_agenda_actions", False)
initial_mml_model_file = params.pop("initial_mml_model_file", None)
params.assert_empty(cls.__name__)
return cls(vocab,
sentence_embedder=sentence_embedder,
action_embedding_dim=action_embedding_dim,
encoder=encoder,
attention_function=attention_function,
beam_size=beam_size,
max_num_finished_states=max_num_finished_states,
dropout=dropout,
max_decoding_steps=max_decoding_steps,
normalize_beam_score_by_length=normalize_beam_score_by_length,
checklist_cost_weight=checklist_cost_weight,
dynamic_cost_weight=dynamic_cost_weight,
penalize_non_agenda_actions=penalize_non_agenda_actions,
initial_mml_model_file=initial_mml_model_file)
def from_params(cls, params: Params) -> 'Attention':
similarity_function = SimilarityFunction.from_params(
params.pop('similarity_function', {}))
normalize = params.pop_bool('normalize', True)
params.assert_empty(cls.__name__)
return cls(similarity_function=similarity_function,
normalize=normalize)
def from_params(cls, vocab, params: Params) -> 'NlvrDirectSemanticParser':
sentence_embedder_params = params.pop("sentence_embedder")
sentence_embedder = TextFieldEmbedder.from_params(
vocab, sentence_embedder_params)
action_embedding_dim = params.pop_int('action_embedding_dim')
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
dropout = params.pop_float('dropout', 0.0)
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(
attention_function_type)
else:
attention_function = None
decoder_beam_search = BeamSearch.from_params(
params.pop("decoder_beam_search"))
max_decoding_steps = params.pop_int("max_decoding_steps")
params.assert_empty(cls.__name__)
return cls(vocab,
sentence_embedder=sentence_embedder,
action_embedding_dim=action_embedding_dim,
encoder=encoder,
attention_function=attention_function,
decoder_beam_search=decoder_beam_search,
max_decoding_steps=max_decoding_steps,
dropout=dropout)
def from_params(cls, vocab, params: Params) -> 'WikiTablesSemanticParser':
question_embedder = TextFieldEmbedder.from_params(vocab, params.pop("question_embedder"))
action_embedding_dim = params.pop_int("action_embedding_dim")
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
entity_encoder = Seq2VecEncoder.from_params(params.pop('entity_encoder'))
max_decoding_steps = params.pop_int("max_decoding_steps")
mixture_feedforward_type = params.pop('mixture_feedforward', None)
if mixture_feedforward_type is not None:
mixture_feedforward = FeedForward.from_params(mixture_feedforward_type)
else:
mixture_feedforward = None
decoder_beam_search = BeamSearch.from_params(params.pop("decoder_beam_search"))
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(attention_function_type)
else:
attention_function = None
dropout = params.pop_float('dropout', 0.0)
num_linking_features = params.pop_int('num_linking_features', 8)
rule_namespace = params.pop('rule_namespace', 'rule_labels')
params.assert_empty(cls.__name__)
return cls(vocab,
question_embedder=question_embedder,
action_embedding_dim=action_embedding_dim,
encoder=encoder,
entity_encoder=entity_encoder,
mixture_feedforward=mixture_feedforward,
decoder_beam_search=decoder_beam_search,
max_decoding_steps=max_decoding_steps,
attention_function=attention_function,
dropout=dropout,
num_linking_features=num_linking_features,
rule_namespace=rule_namespace)
def from_params(cls, vocab, params: Params) -> 'SimpleSeq2SeqCrf':
source_embedder_params = params.pop("source_embedder")
source_embedder = TextFieldEmbedder.from_params(
vocab, source_embedder_params)
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
max_decoding_steps = params.pop("max_decoding_steps")
target_namespace = params.pop("target_namespace", "target_tags")
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(
attention_function_type)
else:
attention_function = None
scheduled_sampling_ratio = params.pop_float("scheduled_sampling_ratio",
0.0)
initializer = InitializerApplicator.from_params(
params.pop('initializer', []))
regularizer = RegularizerApplicator.from_params(
params.pop('regularizer', []))
return cls(vocab,
source_embedder=source_embedder,
encoder=encoder,
max_decoding_steps=max_decoding_steps,
target_namespace=target_namespace,
attention_function=attention_function,
scheduled_sampling_ratio=scheduled_sampling_ratio,
initializer=initializer,
regularizer=regularizer)
def from_params(cls, vocab,
params: Params) -> 'WikiTablesErmSemanticParser':
question_embedder = TextFieldEmbedder.from_params(
vocab, params.pop("question_embedder"))
action_embedding_dim = params.pop_int("action_embedding_dim")
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
entity_encoder = Seq2VecEncoder.from_params(
params.pop('entity_encoder'))
mixture_feedforward_type = params.pop('mixture_feedforward', None)
if mixture_feedforward_type is not None:
mixture_feedforward = FeedForward.from_params(
mixture_feedforward_type)
else:
mixture_feedforward = None
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(
attention_function_type)
else:
attention_function = None
decoder_beam_size = params.pop_int("decoder_beam_size")
decoder_num_finished_states = params.pop_int(
"decoder_num_finished_states", None)
max_decoding_steps = params.pop_int("max_decoding_steps")
normalize_beam_score_by_length = params.pop(
"normalize_beam_score_by_length", False)
use_neighbor_similarity_for_linking = params.pop_bool(
"use_neighbor_similarity_for_linking", False)
dropout = params.pop_float('dropout', 0.0)
num_linking_features = params.pop_int('num_linking_features', 10)
tables_directory = params.pop('tables_directory', '/wikitables/')
rule_namespace = params.pop('rule_namespace', 'rule_labels')
checklist_cost_weight = params.pop_float("checklist_cost_weight", 0.6)
mml_model_file = params.pop('mml_model_file', None)
params.assert_empty(cls.__name__)
return cls(
vocab,
question_embedder=question_embedder,
action_embedding_dim=action_embedding_dim,
encoder=encoder,
entity_encoder=entity_encoder,
mixture_feedforward=mixture_feedforward,
attention_function=attention_function,
decoder_beam_size=decoder_beam_size,
decoder_num_finished_states=decoder_num_finished_states,
max_decoding_steps=max_decoding_steps,
normalize_beam_score_by_length=normalize_beam_score_by_length,
checklist_cost_weight=checklist_cost_weight,
use_neighbor_similarity_for_linking=
use_neighbor_similarity_for_linking,
dropout=dropout,
num_linking_features=num_linking_features,
tables_directory=tables_directory,
rule_namespace=rule_namespace,
initial_mml_model_file=mml_model_file)
def from_params(cls, params: Params) -> 'IntraSentenceAttentionEncoder':
input_dim = params.pop('input_dim')
projection_dim = params.pop('projection_dim', None)
similarity_function = SimilarityFunction.from_params(params.pop('similarity_function', {}))
num_attention_heads = params.pop('num_attention_heads', 1)
combination = params.pop('combination', '1,2')
params.assert_empty(cls.__name__)
return cls(input_dim=input_dim,
projection_dim=projection_dim,
similarity_function=similarity_function,
num_attention_heads=num_attention_heads,
combination=combination)
def from_params(cls, params: Params) -> 'IntraSentenceAttentionEncoder':
input_dim = params.pop_int('input_dim')
projection_dim = params.pop_int('projection_dim', None)
similarity_function = SimilarityFunction.from_params(params.pop('similarity_function', {}))
num_attention_heads = params.pop_int('num_attention_heads', 1)
combination = params.pop('combination', '1,2')
output_dim = params.pop_int('output_dim', None)
params.assert_empty(cls.__name__)
return cls(input_dim=input_dim,
projection_dim=projection_dim,
similarity_function=similarity_function,
num_attention_heads=num_attention_heads,
combination=combination,
output_dim=output_dim)
def from_params(cls, vocab, params: Params) -> 'SimpleSeq2MultiSeq':
tasks = params.pop("tasks")
domains = params.pop("domains")
source_embedder_params = params.pop("source_embedder")
source_embedder = TextFieldEmbedder.from_params(
vocab, source_embedder_params)
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
max_decoding_steps = params.pop("max_decoding_steps")
upos_namespace = params.pop("upos_namespace", "upos_tags")
ner_namespace = params.pop("ner_namespace", "ner_tags")
chunk_namespace = params.pop("chunk_namespace", "chunk_tags")
# upos_namespace = params.pop("upos_namespace")
# ner_namespace = params.pop("ner_namespace")
# chunk_namespace = params.pop("chunk_namespace")
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(
attention_function_type)
else:
attention_function = None
scheduled_sampling_ratio = params.pop_float("scheduled_sampling_ratio",
0.0)
initializer = InitializerApplicator.from_params(
params.pop('initializer', []))
regularizer = RegularizerApplicator.from_params(
params.pop('regularizer', []))
return cls(vocab,
tasks=tasks,
domains=domains,
source_embedder=source_embedder,
encoder=encoder,
max_decoding_steps=max_decoding_steps,
upos_namespace=upos_namespace,
ner_namespace=ner_namespace,
chunk_namespace=chunk_namespace,
attention_function=attention_function,
scheduled_sampling_ratio=scheduled_sampling_ratio,
initializer=initializer,
regularizer=regularizer)
def from_params(cls, vocab, params: Params) -> 'SimpleSeq2Seq':
source_embedder_params = params.pop("source_embedder")
source_embedder = TextFieldEmbedder.from_params(vocab, source_embedder_params)
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
max_decoding_steps = params.pop("max_decoding_steps")
target_namespace = params.pop("target_namespace", "tokens")
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(attention_function_type)
else:
attention_function = None
scheduled_sampling_ratio = params.pop_float("scheduled_sampling_ratio", 0.0)
params.assert_empty(cls.__name__)
return cls(vocab,
source_embedder=source_embedder,
encoder=encoder,
max_decoding_steps=max_decoding_steps,
target_namespace=target_namespace,
attention_function=attention_function,
scheduled_sampling_ratio=scheduled_sampling_ratio)
def from_params(cls, vocab, params: Params) -> 'SpanAe':
source_embedder_params = params.pop("source_embedder")
source_embedder = TextFieldEmbedder.from_params(
vocab, source_embedder_params)
encoder = Seq2SeqEncoder.from_params(params.pop("encoder"))
max_decoding_steps = params.pop("max_decoding_steps")
target_namespace = params.pop("target_namespace", "tokens")
# If no attention function is specified, we should not use attention, not attention with
# default similarity function.
attention_function_type = params.pop("attention_function", None)
if attention_function_type is not None:
attention_function = SimilarityFunction.from_params(
attention_function_type)
else:
attention_function = None
scheduled_sampling_ratio = params.pop_float("scheduled_sampling_ratio",
0.0)
spans_extractor, spans_scorer_feedforward = None, None
spans_extractor_params = params.pop("span_extractor", None)
if spans_extractor_params is not None:
spans_extractor = SpanExtractor.from_params(spans_extractor_params)
spans_scorer_params = params.pop("span_scorer_feedforward", None)
if spans_scorer_params is not None:
spans_scorer_feedforward = FeedForward.from_params(
spans_scorer_params)
spans_per_word = params.pop_float("spans_per_word")
params.assert_empty(cls.__name__)
return cls(vocab,
source_embedder=source_embedder,
encoder=encoder,
max_decoding_steps=max_decoding_steps,
spans_per_word=spans_per_word,
target_namespace=target_namespace,
attention_function=attention_function,
scheduled_sampling_ratio=scheduled_sampling_ratio,
spans_extractor=spans_extractor,
spans_scorer_feedforward=spans_scorer_feedforward)
def test_registry_has_builtin_similarity_functions(self):
assert SimilarityFunction.by_name("dot_product").__name__ == 'DotProductSimilarity'
assert SimilarityFunction.by_name("bilinear").__name__ == 'BilinearSimilarity'
assert SimilarityFunction.by_name("linear").__name__ == 'LinearSimilarity'
assert SimilarityFunction.by_name("cosine").__name__ == 'CosineSimilarity'
def from_params(cls, params: Params) -> 'Attention':
similarity_function = SimilarityFunction.from_params(params.pop('similarity_function', {}))
normalize = params.pop_bool('normalize', True)
params.assert_empty(cls.__name__)
return cls(similarity_function=similarity_function,
normalize=normalize)
def from_params(cls, params: Params) -> 'MatrixAttention':
similarity_function = SimilarityFunction.from_params(params.pop('similarity_function', {}))
params.assert_empty(cls.__name__)
return cls(similarity_function=similarity_function)
def from_params(cls, params: Params) -> 'MatrixAttention':
similarity_function = SimilarityFunction.from_params(
params.pop('similarity_function', {}))
return cls(similarity_function=similarity_function)
