def test_smoke_attention_charts():
"""A smoke test to make sure no errors pop up while making attention charts"""
model_folder = "tests/demo/models"
model_path = list(pathlib.Path(model_folder).glob("*"))[0]
interpreter = load_interpreter(model_folder, str(model_path.parts[-1]))
text_input = "hello world this is dog"
diag_data, tokens = fetch_attention_feats(interpreter, text_input)
_ = make_attention_charts(diag_data, tokens)
def test_can_pick_up_entities():
model_folder = "tests/demo/models"
model_path = list(pathlib.Path(model_folder).glob("*"))[0]
interpreter = load_interpreter(model_folder, str(model_path.parts[-1]))
blob, nlu_dict, tokens = fetch_info_from_message(interpreter=interpreter,
text_input="hello python")
assert blob["entities"][0]["entity"] == "proglang"
assert blob["entities"][0]["value"] == "python"
def test_load_interpreter_can_load():
model_folder = "tests/demo/models"
model_path = list(pathlib.Path(model_folder).glob("*"))[0]
interpreter = load_interpreter(model_folder, str(model_path.parts[-1]))
blob, nlu_dict, tokens = fetch_info_from_message(interpreter=interpreter,
text_input="hello world")
assert blob["text"] == "hello world"
assert tokens == ["hello", "world"]
all_intents = [
"talk_code",
"bot_challenge",
"mood_unhappy",
"mood_great",
"deny",
"affirm",
"goodbye",
"greet",
]
assert [i["name"] in all_intents for i in blob["intent_ranking"]]
def test_load_interpreter_can_load():
model_folder = "tests/demo/models"
model_path = list(pathlib.Path(model_folder).glob("*"))[0]
interpreter = load_interpreter(model_folder, str(model_path.parts[-1]))
msg = Message("hello world")
for i, element in enumerate(interpreter.pipeline):
element.process(msg)
nlu_dict = msg.as_dict_nlu()
assert nlu_dict["text"] == "hello world"
assert [t.text
for t in nlu_dict["tokens"]] == ["hello", "world", "__CLS__"]
all_intents = [
"talk_code",
"bot_challenge",
"mood_unhappy",
"mood_great",
"deny",
"affirm",
"goodbye",
"greet",
]
assert [i["name"] in all_intents for i in nlu_dict["intent_ranking"]]
parser = argparse.ArgumentParser(description="")
parser.add_argument("--folder", help="Pass the model folder.")
args = parser.parse_args()
model_folder = args.folder
st.markdown("# Arabic KYC NLU Model Summary")
st.markdown("You can select a model on the left to interact with.")
model_files = [
str(p.parts[-1]) for p in pathlib.Path(model_folder).glob("*.tar.gz")
]
model_file = st.sidebar.selectbox("What model do you want to use", model_files)
interpreter = load_interpreter(model_folder, model_file)
text_input = st.text_input("Text Input for Model", "Hello")
blob, nlu_dict, tokens = fetch_info_from_message(interpreter=interpreter,
text_input=text_input)
st.markdown("## Tokens and Entities")
st.write(
create_displacy_chart(tokens=tokens, entities=blob["entities"]),
unsafe_allow_html=True,
)
st.markdown("## Intents")
chart_data = pd.DataFrame(blob["intent_ranking"]).sort_values("name")
