def test_fingerprint_is_same_when_loading_data_again():
from rasa.shared.importers.utils import training_data_from_paths
files = [
"data/examples/rasa/demo-rasa.yml",
"data/examples/rasa/demo-rasa-responses.yml",
]
td1 = training_data_from_paths(files, language="en")
td2 = training_data_from_paths(files, language="en")
assert td1.fingerprint() == td2.fingerprint()
def test_demo_data(files: List[Text]):
from rasa.shared.importers.utils import training_data_from_paths
trainingdata = training_data_from_paths(files, language="en")
assert trainingdata.intents == {
"affirm",
"greet",
"restaurant_search",
"goodbye",
"chitchat",
}
assert trainingdata.entities == {"location", "cuisine"}
assert set(trainingdata.responses.keys()) == {
"utter_chitchat/ask_name",
"utter_chitchat/ask_weather",
}
assert len(trainingdata.training_examples) == 46
assert len(trainingdata.intent_examples) == 46
assert len(trainingdata.response_examples) == 4
assert len(trainingdata.entity_examples) == 11
assert len(trainingdata.responses) == 2
assert trainingdata.entity_synonyms == {
"Chines": "chinese",
"Chinese": "chinese",
"chines": "chinese",
"vegg": "vegetarian",
"veggie": "vegetarian",
}
assert trainingdata.regex_features == [
{"name": "greet", "pattern": r"hey[^\s]*"},
{"name": "zipcode", "pattern": r"[0-9]{5}"},
]
def test_train_test_split(filepaths: List[Text]):
from rasa.shared.importers.utils import training_data_from_paths
training_data = training_data_from_paths(filepaths, language="en")
assert training_data.intents == {
"affirm",
"greet",
"restaurant_search",
"goodbye",
"chitchat",
}
assert training_data.entities == {"location", "cuisine"}
assert set(training_data.responses.keys()) == {
"utter_chitchat/ask_name",
"utter_chitchat/ask_weather",
}
NUM_TRAIN_EXAMPLES = 46
NUM_RESPONSE_EXAMPLES = 4
assert len(training_data.training_examples) == NUM_TRAIN_EXAMPLES
assert len(training_data.intent_examples) == NUM_TRAIN_EXAMPLES
assert len(training_data.response_examples) == NUM_RESPONSE_EXAMPLES
for train_percent in range(50, 95, 5):
train_frac = train_percent / 100.0
train_split, test_split = training_data.train_test_split(train_frac)
assert (len(test_split.training_examples) +
len(train_split.training_examples) == NUM_TRAIN_EXAMPLES)
num_classes = (
len(training_data.number_of_examples_per_intent.keys()) +
-len(training_data.retrieval_intents) +
len(training_data.number_of_examples_per_response))
expected_num_train_examples_floor = int(train_frac *
NUM_TRAIN_EXAMPLES)
if NUM_TRAIN_EXAMPLES - expected_num_train_examples_floor < num_classes:
expected_num_train_examples_floor = NUM_TRAIN_EXAMPLES - num_classes - 1
assert len(
train_split.training_examples) >= expected_num_train_examples_floor
assert (len(train_split.training_examples) <=
expected_num_train_examples_floor + 1)
assert len(training_data.number_of_examples_per_intent.keys()) == len(
test_split.number_of_examples_per_intent.keys())
assert len(training_data.number_of_examples_per_intent.keys()) == len(
train_split.number_of_examples_per_intent.keys())
assert len(
training_data.number_of_examples_per_response.keys()) == len(
train_split.number_of_examples_per_response.keys())
assert len(
training_data.number_of_examples_per_response.keys()) == len(
train_split.number_of_examples_per_response.keys())
def test_fingerprint_is_different_when_lookup_table_has_changed(
monkeypatch: MonkeyPatch,
):
from rasa.shared.importers.utils import training_data_from_paths
files = [
"data/test/lookup_tables/lookup_table.json",
]
td1 = training_data_from_paths(files, language="en")
fingerprint1 = td1.fingerprint()
monkeypatch.setattr(
TrainingData,
"_load_lookup_table",
Mock(return_value={"name": "plates", "elements": "tacos\nbeef"}),
)
td2 = training_data_from_paths(files, language="en")
fingerprint2 = td2.fingerprint()
assert fingerprint1 != fingerprint2
def test_training_data_fingerprint_incorporates_tokens(
whitespace_tokenizer: WhitespaceTokenizer, ):
from rasa.shared.importers.utils import training_data_from_paths
files = [
"data/examples/rasa/demo-rasa.yml",
"data/examples/rasa/demo-rasa-responses.yml",
]
training_data = training_data_from_paths(files, language="en")
fp1 = training_data.fingerprint()
whitespace_tokenizer.process_training_data(training_data)
# training data fingerprint has changed
assert fp1 != training_data.fingerprint()
def test_train_test_split_with_random_seed(filepaths):
from rasa.shared.importers.utils import training_data_from_paths
td = training_data_from_paths(filepaths, language="en")
td_train_1, td_test_1 = td.train_test_split(train_frac=0.8, random_seed=1)
td_train_2, td_test_2 = td.train_test_split(train_frac=0.8, random_seed=1)
train_1_intent_examples = [e.get(TEXT) for e in td_train_1.intent_examples]
train_2_intent_examples = [e.get(TEXT) for e in td_train_2.intent_examples]
test_1_intent_examples = [e.get(TEXT) for e in td_test_1.intent_examples]
test_2_intent_examples = [e.get(TEXT) for e in td_test_2.intent_examples]
assert train_1_intent_examples == train_2_intent_examples
assert test_1_intent_examples == test_2_intent_examples
def test_training_data_fingerprint_incorporates_features():
from rasa.shared.importers.utils import training_data_from_paths
files = [
"data/examples/rasa/demo-rasa.yml",
"data/examples/rasa/demo-rasa-responses.yml",
]
training_data = training_data_from_paths(files, language="en")
fp1 = training_data.fingerprint()
big_array = np.random.random((128, 128))
f1 = Features(big_array, FEATURE_TYPE_SENTENCE, TEXT, "RegexFeaturizer")
training_data.training_examples[0].add_features(f1)
# training data fingerprint has changed
assert fp1 != training_data.fingerprint()
def test_demo_data_filter_out_retrieval_intents(files):
from rasa.shared.importers.utils import training_data_from_paths
training_data = training_data_from_paths(files, language="en")
assert len(training_data.training_examples) == 46
training_data_filtered = training_data.filter_training_examples(
lambda ex: ex.get(INTENT_RESPONSE_KEY) is None)
assert len(training_data_filtered.training_examples) == 42
training_data_filtered_2 = training_data.filter_training_examples(
lambda ex: ex.get(INTENT_RESPONSE_KEY) is not None)
assert len(training_data_filtered_2.training_examples) == 4
# make sure filtering operation doesn't mutate the source training data
assert len(training_data.training_examples) == 46
async def get_nlu_data(self, languages=True) -> Dict[Text, TrainingData]:
language = None
if isinstance(languages, str):
language = languages
languages = [language]
if not isinstance(languages, list):
languages = self.nlu_config.keys()
td = {}
for lang in languages:
try:
td[lang] = utils.training_data_from_paths(
self.path_for_nlu_lang(lang),
lang,
)
except ValueError as e:
if str(e).startswith("Unknown data format"):
td[lang] = TrainingData()
if language: return td.get(language, TrainingData())
return td
def test_train_test_split(filepaths: List[Text]):
from rasa.shared.importers.utils import training_data_from_paths
trainingdata = training_data_from_paths(filepaths, language="en")
assert trainingdata.intents == {
"affirm",
"greet",
"restaurant_search",
"goodbye",
"chitchat",
}
assert trainingdata.entities == {"location", "cuisine"}
assert set(trainingdata.responses.keys()) == {
"utter_chitchat/ask_name",
"utter_chitchat/ask_weather",
}
assert len(trainingdata.training_examples) == 46
assert len(trainingdata.intent_examples) == 46
assert len(trainingdata.response_examples) == 4
trainingdata_train, trainingdata_test = trainingdata.train_test_split(
train_frac=0.8)
assert (len(trainingdata_test.training_examples) +
len(trainingdata_train.training_examples) == 46)
assert len(trainingdata_train.training_examples) == 34
assert len(trainingdata_test.training_examples) == 12
assert len(trainingdata.number_of_examples_per_intent.keys()) == len(
trainingdata_test.number_of_examples_per_intent.keys())
assert len(trainingdata.number_of_examples_per_intent.keys()) == len(
trainingdata_train.number_of_examples_per_intent.keys())
assert len(trainingdata.number_of_examples_per_response.keys()) == len(
trainingdata_test.number_of_examples_per_response.keys())
assert len(trainingdata.number_of_examples_per_response.keys()) == len(
trainingdata_train.number_of_examples_per_response.keys())
def get_nlu_data(self, language: Optional[Text] = "en") -> TrainingData:
"""Retrieves NLU training data (see parent class for full docstring)."""
return utils.training_data_from_paths(self._nlu_files, language)
async def get_nlu_data(self,
language: Optional[Text] = "en") -> TrainingData:
return utils.training_data_from_paths(self._nlu_files, language)
