Update app.py
Browse files
app.py
CHANGED
@@ -5,7 +5,6 @@ import pandas as pd
|
|
5 |
import os
|
6 |
import gradio as gr
|
7 |
|
8 |
-
# Load the models and tokenizers
|
9 |
bert_model_name = 'bert-base-uncased'
|
10 |
bert_tokenizer = BertTokenizer.from_pretrained(bert_model_name)
|
11 |
bert_model = BertModel.from_pretrained(bert_model_name)
|
@@ -14,7 +13,6 @@ gpt2_model_name = 'gpt2'
|
|
14 |
gpt2_tokenizer = GPT2Tokenizer.from_pretrained(gpt2_model_name)
|
15 |
gpt2_model = GPT2LMHeadModel.from_pretrained(gpt2_model_name)
|
16 |
|
17 |
-
# Load the data
|
18 |
data = {
|
19 |
"questions": [
|
20 |
"What is Rookus?",
|
@@ -39,6 +37,18 @@ data = {
|
|
39 |
"default_answers": "I'm sorry, I cannot answer this right now. Your question has been saved, and we will get back to you with a response soon."
|
40 |
}
|
41 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
def get_bert_embeddings(texts):
|
43 |
inputs = bert_tokenizer(texts, return_tensors='pt', padding=True, truncation=True)
|
44 |
with torch.no_grad():
|
@@ -70,18 +80,10 @@ def answer_query(user_query):
|
|
70 |
answer_index = data['questions'].index(closest_question)
|
71 |
answer = data['answers'][answer_index]
|
72 |
else:
|
73 |
-
excel_file = 'new_questions1.xlsx'
|
74 |
-
if not os.path.isfile(excel_file):
|
75 |
-
df = pd.DataFrame(columns=['question'])
|
76 |
-
df.to_excel(excel_file, index=False)
|
77 |
-
|
78 |
-
new_data = pd.DataFrame({'questions': [user_query]})
|
79 |
-
df = pd.read_excel(excel_file)
|
80 |
-
df = pd.concat([df, new_data], ignore_index=True)
|
81 |
-
with pd.ExcelWriter(excel_file, engine='openpyxl', mode='w') as writer:
|
82 |
-
df.to_excel(writer, index=False)
|
83 |
answer = data['default_answers']
|
84 |
-
|
|
|
|
|
85 |
return answer
|
86 |
|
87 |
iface = gr.Interface(
|
|
|
5 |
import os
|
6 |
import gradio as gr
|
7 |
|
|
|
8 |
bert_model_name = 'bert-base-uncased'
|
9 |
bert_tokenizer = BertTokenizer.from_pretrained(bert_model_name)
|
10 |
bert_model = BertModel.from_pretrained(bert_model_name)
|
|
|
13 |
gpt2_tokenizer = GPT2Tokenizer.from_pretrained(gpt2_model_name)
|
14 |
gpt2_model = GPT2LMHeadModel.from_pretrained(gpt2_model_name)
|
15 |
|
|
|
16 |
data = {
|
17 |
"questions": [
|
18 |
"What is Rookus?",
|
|
|
37 |
"default_answers": "I'm sorry, I cannot answer this right now. Your question has been saved, and we will get back to you with a response soon."
|
38 |
}
|
39 |
|
40 |
+
def save_to_excel(user_query, default_answer, excel_file='new_prompts.xlsx'):
|
41 |
+
if not os.path.isfile(excel_file):
|
42 |
+
df = pd.DataFrame(columns=['question', 'default_answer'])
|
43 |
+
else:
|
44 |
+
df = pd.read_excel(excel_file)
|
45 |
+
|
46 |
+
new_data = pd.DataFrame({'question': [user_query], 'default_answer': [default_answer]})
|
47 |
+
df = pd.concat([df, new_data], ignore_index=True)
|
48 |
+
|
49 |
+
with pd.ExcelWriter(excel_file, engine='openpyxl', mode='w') as writer:
|
50 |
+
df.to_excel(writer, index=False)
|
51 |
+
|
52 |
def get_bert_embeddings(texts):
|
53 |
inputs = bert_tokenizer(texts, return_tensors='pt', padding=True, truncation=True)
|
54 |
with torch.no_grad():
|
|
|
80 |
answer_index = data['questions'].index(closest_question)
|
81 |
answer = data['answers'][answer_index]
|
82 |
else:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
83 |
answer = data['default_answers']
|
84 |
+
|
85 |
+
save_to_excel(user_query, answer)
|
86 |
+
|
87 |
return answer
|
88 |
|
89 |
iface = gr.Interface(
|