Update src/streamlit_app.py
Browse files- src/streamlit_app.py +40 -1
src/streamlit_app.py
CHANGED
|
@@ -16,6 +16,10 @@ def instantiate_roberta(top_k : int, text : str) -> dict:
|
|
| 16 |
pipe = pipeline("fill-mask", model="Iscte-Sintra/RoBERTa-Kriolu", tokenizer="Iscte-Sintra/RoBERTa-Kriolu", token=token)
|
| 17 |
return pipe(text, top_k=top_k)
|
| 18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
|
| 20 |
def build_gpt2_page():
|
| 21 |
try:
|
|
@@ -66,13 +70,48 @@ def build_roberta_page():
|
|
| 66 |
else:
|
| 67 |
predicted_text = st.text_input("Predicted Token", disabled=True)
|
| 68 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 69 |
# Your dictionary of models
|
| 70 |
-
model_dict = {'RoBERTa': 1, 'GPT-2':
|
| 71 |
|
| 72 |
# Always appears at the top of the sidebar
|
| 73 |
selected_model = st.sidebar.selectbox("Architecture", list(model_dict.keys()))
|
| 74 |
|
| 75 |
if model_dict[selected_model] == 1:
|
| 76 |
build_roberta_page()
|
|
|
|
|
|
|
| 77 |
else:
|
| 78 |
build_gpt2_page()
|
|
|
|
| 16 |
pipe = pipeline("fill-mask", model="Iscte-Sintra/RoBERTa-Kriolu", tokenizer="Iscte-Sintra/RoBERTa-Kriolu", token=token)
|
| 17 |
return pipe(text, top_k=top_k)
|
| 18 |
|
| 19 |
+
def instatiate_albertina(top_k : int, text : str) -> dict:
|
| 20 |
+
pipe = pipeline("fill-mask", model="Iscte-Sintra/Albertina-Kriolu", tokenizer="Iscte-Sintra/Albertina-Kriolu", token=token)
|
| 21 |
+
return pipe(text, top_k=top_k)
|
| 22 |
+
|
| 23 |
|
| 24 |
def build_gpt2_page():
|
| 25 |
try:
|
|
|
|
| 70 |
else:
|
| 71 |
predicted_text = st.text_input("Predicted Token", disabled=True)
|
| 72 |
|
| 73 |
+
def build_albertina_page():
|
| 74 |
+
st.title("Albertina : Encoder")
|
| 75 |
+
|
| 76 |
+
top_k = st.sidebar.number_input('Number of predictions to return', min_value=1, max_value=5, value=1, step=1)
|
| 77 |
+
|
| 78 |
+
st.write("Enter a sentence with a **<mask>** token, and the model will predict the missing word.")
|
| 79 |
+
|
| 80 |
+
results = None
|
| 81 |
+
|
| 82 |
+
col1, col2 = st.columns(2)
|
| 83 |
+
|
| 84 |
+
with col1:
|
| 85 |
+
st.subheader("Input")
|
| 86 |
+
input_text = st.text_input("Input Sentence", "Katxor sta trás di <mask>.")
|
| 87 |
+
|
| 88 |
+
submit = st.button("Submit")
|
| 89 |
+
try:
|
| 90 |
+
if submit and input_text:
|
| 91 |
+
results = instantiate_roberta(top_k, input_text)
|
| 92 |
+
except Exception as e:
|
| 93 |
+
st.warning('There must be a special token "<mask>" in sentence!', icon="⚠️")
|
| 94 |
+
st.warning(e)
|
| 95 |
+
|
| 96 |
+
with col2:
|
| 97 |
+
st.subheader("Prediction")
|
| 98 |
+
if results:
|
| 99 |
+
predicted_text = st.text_input("Predicted Token", value=results[0]['sequence'], disabled=True)
|
| 100 |
+
for result in results:
|
| 101 |
+
st.write(f"**Prediction**: {result['token_str']} | **Confidence**: {round(result['score'], 4)}")
|
| 102 |
+
else:
|
| 103 |
+
predicted_text = st.text_input("Predicted Token", disabled=True)
|
| 104 |
+
|
| 105 |
+
|
| 106 |
# Your dictionary of models
|
| 107 |
+
model_dict = {'RoBERTa': 1,'Albertina':2 ,'GPT-2': 3}
|
| 108 |
|
| 109 |
# Always appears at the top of the sidebar
|
| 110 |
selected_model = st.sidebar.selectbox("Architecture", list(model_dict.keys()))
|
| 111 |
|
| 112 |
if model_dict[selected_model] == 1:
|
| 113 |
build_roberta_page()
|
| 114 |
+
elif model_dict[selected_model] == 2:
|
| 115 |
+
build_albertina_page()
|
| 116 |
else:
|
| 117 |
build_gpt2_page()
|