def common_logic(cls, cur_name):
name = cur_name.split("test_")[1]
for spec, sent_ in cls.out[name].items():
sent = [v.copy() for v in sent_]
add_basic_edges(sent)
converted, _ = convert(
[sent],
True,
True,
True,
math.inf,
False,
False,
False,
False,
False,
funcs_to_cancel=list(
set(api.get_conversion_names()).difference({
name, "extra_inner_weak_modifier_verb_reconstruction"
})))
serialized_conllu = serialize_conllu(converted, [None], False)
for gold_line, out_line in zip(cls.gold[name][spec],
serialized_conllu.split("\n")):
assert out_line.split() == gold_line, spec + str(
print("\n")) + str(
[print(s) for s in serialized_conllu.split("\n")])
def common_logic_combined(cls, cur_name, rnac=False):
name = cur_name.split("test_combined_")[1]
for spec, sent_ in cls.out[name].items():
sent = {k: v.copy() for k, v in sent_.items()}
add_basic_edges(sent)
converted, _ = convert([sent], True, True, True, math.inf, False,
False, rnac, False, False, ConvsCanceler())
serialized_conllu = serialize_conllu(converted, [None], False)
for gold_line, out_line in zip(cls.gold_combined[name][spec],
serialized_conllu.split("\n")):
assert gold_line == out_line.split(), spec + str(
[print(s) for s in serialized_conllu.split("\n")])
def test_no_node_adding(self):
self.common_logic_combined("test_combined_no_node_adding", rnac=True)
for cur_func_name in api.get_conversion_names():
if cur_func_name in ['extra_inner_weak_modifier_verb_reconstruction']:
continue
test_func_name = "test_" + cur_func_name
setattr(
TestConversions, test_func_name,
staticmethod(lambda func_name=test_func_name: TestConversions.
common_logic(func_name)))
combined_func_name = "test_combined_" + cur_func_name
setattr(
TestConversions, combined_func_name,
staticmethod(lambda func_name=combined_func_name: TestConversions.
common_logic_combined(func_name)))
type=str,
default=None)
args = arg_parser.parse_args()
# what group of stratagies to use
if args.strat_start >= 0:
if args.strat_end >= 0:
strategies = strategies[args.strat_start:args.strat_end]
else:
strategies = strategies[args.strat_start:args.strat_start + 1]
# for ablation test, get ablated function name
ablation = ""
if args.ablation >= 0:
ablations = uda_api.get_conversion_names()
ablation = sorted(list(ablations))[args.ablation]
# if we want to use specific triggers (w/o we use all the triggers under triggers folder)
if args.specified_triggers:
g_triggers = args.specified_triggers
# the different actions we take:
# annotate: as we need to have annotated dataset for Spike to Index.
# After running this command, we need to Index the annotated data using Spike,
# only then we should run the 'generate' and 'eval' commands.
# generate: as we need to generate patterns (from the train set)
# that would later be attested on the dev and test sets using the 'eval' command
# eval: as we want to filter and evaluate the generated patterns on the dev and test sets (respectively).
# ablations: a different test, to check each conversions' contribution.
if args.action == 'annotate':