Spaces:
Runtime error
Runtime error
mlkorra
commited on
Commit
•
33ae6dd
1
Parent(s):
decd5de
update app
Browse files- app.py +34 -16
- mlm_custom/mlm_test_config.csv +1 -1
app.py
CHANGED
@@ -5,16 +5,18 @@ from transformers import AutoTokenizer,AutoModelForMaskedLM
|
|
5 |
from transformers import pipeline
|
6 |
import os
|
7 |
import json
|
|
|
8 |
|
9 |
@st.cache(show_spinner=False,persist=True)
|
10 |
def load_model(masked_text,model_name):
|
11 |
|
12 |
model = AutoModelForMaskedLM.from_pretrained(model_name, from_flax=True)
|
13 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
14 |
-
# tokenizer.save_pretrained('exported_pytorch_model')
|
15 |
-
# model.save_pretrained('exported_pytorch_model')
|
16 |
nlp = pipeline('fill-mask', model=model, tokenizer=tokenizer)
|
17 |
-
|
|
|
|
|
|
|
18 |
result_sentence = nlp(masked_text)
|
19 |
|
20 |
return result_sentence[0]['sequence']
|
@@ -27,7 +29,7 @@ def main():
|
|
27 |
)
|
28 |
|
29 |
models = st.multiselect(
|
30 |
-
"Choose models",
|
31 |
['flax-community/roberta-hindi','mrm8488/HindiBERTa','ai4bharat/indic-bert',\
|
32 |
'neuralspace-reverie/indic-transformers-hi-bert',
|
33 |
'surajp/RoBERTa-hindi-guj-san'],
|
@@ -40,24 +42,40 @@ def main():
|
|
40 |
texts = target_text_df['text']
|
41 |
|
42 |
st.sidebar.title("Hindi MLM")
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
st.write('You selected:', masked_text)
|
47 |
|
48 |
results_df = pd.DataFrame(columns = ['Model Name','Masked Text','Filled Masked Text'])
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
49 |
|
50 |
-
|
|
|
|
|
|
|
51 |
|
52 |
-
if st.button('Fill the Mask!'):
|
53 |
-
with st.spinner("Filling the Mask..."):
|
54 |
filled_sentence = load_model(masked_text,selected_model)
|
55 |
-
|
56 |
-
|
57 |
-
|
58 |
|
59 |
-
|
|
|
|
|
60 |
|
61 |
-
|
|
|
62 |
if __name__ == "__main__":
|
63 |
main()
|
|
|
5 |
from transformers import pipeline
|
6 |
import os
|
7 |
import json
|
8 |
+
import random
|
9 |
|
10 |
@st.cache(show_spinner=False,persist=True)
|
11 |
def load_model(masked_text,model_name):
|
12 |
|
13 |
model = AutoModelForMaskedLM.from_pretrained(model_name, from_flax=True)
|
14 |
tokenizer = AutoTokenizer.from_pretrained(model_name)
|
|
|
|
|
15 |
nlp = pipeline('fill-mask', model=model, tokenizer=tokenizer)
|
16 |
+
|
17 |
+
MASK_TOKEN = tokenizer.mask_token
|
18 |
+
|
19 |
+
masked_text = masked_text.replace("<mask>",MASK_TOKEN)
|
20 |
result_sentence = nlp(masked_text)
|
21 |
|
22 |
return result_sentence[0]['sequence']
|
|
|
29 |
)
|
30 |
|
31 |
models = st.multiselect(
|
32 |
+
"Choose models",
|
33 |
['flax-community/roberta-hindi','mrm8488/HindiBERTa','ai4bharat/indic-bert',\
|
34 |
'neuralspace-reverie/indic-transformers-hi-bert',
|
35 |
'surajp/RoBERTa-hindi-guj-san'],
|
|
|
42 |
texts = target_text_df['text']
|
43 |
|
44 |
st.sidebar.title("Hindi MLM")
|
45 |
+
|
46 |
+
pick_random = st.sidebar.checkbox("Pick any random text")
|
47 |
+
|
48 |
+
#st.write('You selected:', masked_text)
|
49 |
|
50 |
results_df = pd.DataFrame(columns = ['Model Name','Masked Text','Filled Masked Text'])
|
51 |
+
|
52 |
+
model_names = []
|
53 |
+
masked_texts = []
|
54 |
+
filled_masked_texts = []
|
55 |
+
|
56 |
+
if pick_random:
|
57 |
+
random_text = texts[random.randint(0,texts.shape[0]-1)]
|
58 |
+
masked_text = st.text_area("Please type a masked sentence to fill",random_text)
|
59 |
+
else:
|
60 |
+
select_text = st.sidebar.selectbox('Select any of the following text',\
|
61 |
+
texts)
|
62 |
+
masked_text = st.text_area("Please type a masked sentence to fill",select_text)
|
63 |
|
64 |
+
if st.button('Fill the Mask!'):
|
65 |
+
with st.spinner("Filling the Mask..."):
|
66 |
+
|
67 |
+
for selected_model in models:
|
68 |
|
|
|
|
|
69 |
filled_sentence = load_model(masked_text,selected_model)
|
70 |
+
model_names.append(selected_model)
|
71 |
+
masked_texts.append(masked_text)
|
72 |
+
filled_masked_texts.append(filled_sentence)
|
73 |
|
74 |
+
results_df['Model Name'] = model_names
|
75 |
+
results_df['Masked Text'] = masked_texts
|
76 |
+
results_df['Filled Masked Text'] = filled_masked_texts
|
77 |
|
78 |
+
st.table(results_df)
|
79 |
+
|
80 |
if __name__ == "__main__":
|
81 |
main()
|
mlm_custom/mlm_test_config.csv
CHANGED
@@ -3,4 +3,4 @@ flax-community/roberta-hindi,,,TRUE,TRUE,TRUE
|
|
3 |
mrm8488/HindiBERTa,,,FALSE,TRUE,TRUE
|
4 |
ai4bharat/indic-bert,,,FALSE,FALSE,FALSE
|
5 |
neuralspace-reverie/indic-transformers-hi-bert,,,FALSE,TRUE,TRUE
|
6 |
-
surajp/RoBERTa-hindi-guj-san,,,FALSE,TRUE,TRUE
|
|
|
3 |
mrm8488/HindiBERTa,,,FALSE,TRUE,TRUE
|
4 |
ai4bharat/indic-bert,,,FALSE,FALSE,FALSE
|
5 |
neuralspace-reverie/indic-transformers-hi-bert,,,FALSE,TRUE,TRUE
|
6 |
+
surajp/RoBERTa-hindi-guj-san,,,FALSE,TRUE,TRUE
|