Spaces:
Runtime error
Runtime error
Delete app.py
Browse files
app.py
DELETED
@@ -1,138 +0,0 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
from sentence_transformers import SentenceTransformer, util
|
3 |
-
from transformers import pipeline, GPT2Tokenizer
|
4 |
-
import os
|
5 |
-
|
6 |
-
# Define paths and model identifiers for easy reference and maintenance
|
7 |
-
filename = "output_country_details.txt" # Filename for stored country details
|
8 |
-
retrieval_model_name = 'output/sentence-transformer-finetuned/'
|
9 |
-
gpt2_model_name = "gpt2" # Identifier for the GPT-2 model used
|
10 |
-
tokenizer = GPT2Tokenizer.from_pretrained(gpt2_model_name)
|
11 |
-
|
12 |
-
# Load models and handle potential failures gracefully
|
13 |
-
try:
|
14 |
-
retrieval_model = SentenceTransformer(retrieval_model_name)
|
15 |
-
gpt_model = pipeline("text-generation", model=gpt2_model_name)
|
16 |
-
print("Models loaded successfully.")
|
17 |
-
except Exception as e:
|
18 |
-
print(f"Failed to load models: {e}")
|
19 |
-
|
20 |
-
def load_and_preprocess_text(filename):
|
21 |
-
"""
|
22 |
-
Load text data from a file and preprocess it by stripping whitespace and ignoring empty lines.
|
23 |
-
|
24 |
-
Args:
|
25 |
-
filename (str): Path to the file containing text data.
|
26 |
-
|
27 |
-
Returns:
|
28 |
-
list of str: Preprocessed lines of text from the file.
|
29 |
-
"""
|
30 |
-
try:
|
31 |
-
with open(filename, 'r', encoding='utf-8') as file:
|
32 |
-
segments = [line.strip() for line in file if line.strip()]
|
33 |
-
print("Text loaded and preprocessed successfully.")
|
34 |
-
return segments
|
35 |
-
except Exception as e:
|
36 |
-
print(f"Failed to load or preprocess text: {e}")
|
37 |
-
return []
|
38 |
-
|
39 |
-
segments = load_and_preprocess_text(filename)
|
40 |
-
|
41 |
-
def find_relevant_segment(user_query, segments):
|
42 |
-
"""
|
43 |
-
Identify the most relevant text segment from a list based on a user's query using sentence embeddings.
|
44 |
-
|
45 |
-
Args:
|
46 |
-
user_query (str): User's input query.
|
47 |
-
segments (list of str): List of text segments to search from.
|
48 |
-
|
49 |
-
Returns:
|
50 |
-
str: The text segment that best matches the query.
|
51 |
-
"""
|
52 |
-
try:
|
53 |
-
query_embedding = retrieval_model.encode(user_query)
|
54 |
-
segment_embeddings = retrieval_model.encode(segments)
|
55 |
-
similarities = util.pytorch_cos_sim(query_embedding, segment_embeddings)[0]
|
56 |
-
best_idx = similarities.argmax()
|
57 |
-
print("Relevant segment found:", segments[best_idx])
|
58 |
-
return segments[best_idx]
|
59 |
-
except Exception as e:
|
60 |
-
print(f"Error finding relevant segment: {e}")
|
61 |
-
return ""
|
62 |
-
|
63 |
-
def generate_response(user_query, relevant_segment):
|
64 |
-
"""
|
65 |
-
Generate a response to a user's query using a text generation model based on a relevant text segment.
|
66 |
-
|
67 |
-
Args:
|
68 |
-
user_query (str): The user's query.
|
69 |
-
relevant_segment (str): The segment of text that is relevant to the query.
|
70 |
-
|
71 |
-
Returns:
|
72 |
-
str: A response generated from the model.
|
73 |
-
"""
|
74 |
-
try:
|
75 |
-
prompt = f"Thank you for your question! This is an additional fact about your topic: {relevant_segment}"
|
76 |
-
max_tokens = len(tokenizer(prompt)['input_ids']) + 50
|
77 |
-
response = gpt_model(prompt, max_length=max_tokens, temperature=0.25)[0]['generated_text']
|
78 |
-
response_cleaned = clean_up_response(response, relevant_segment)
|
79 |
-
return response_cleaned
|
80 |
-
except Exception as e:
|
81 |
-
print(f"Error generating response: {e}")
|
82 |
-
return ""
|
83 |
-
|
84 |
-
def clean_up_response(response, segments):
|
85 |
-
"""
|
86 |
-
Clean and format the generated response by removing empty sentences and repetitive parts.
|
87 |
-
|
88 |
-
Args:
|
89 |
-
response (str): The raw response generated by the model.
|
90 |
-
segments (str): The text segment used to generate the response.
|
91 |
-
|
92 |
-
Returns:
|
93 |
-
str: Cleaned and formatted response.
|
94 |
-
"""
|
95 |
-
sentences = response.split('.')
|
96 |
-
cleaned_sentences = []
|
97 |
-
for sentence in sentences:
|
98 |
-
if sentence.strip() and sentence.strip() not in segments and sentence.strip() not in cleaned_sentences:
|
99 |
-
cleaned_sentences.append(sentence.strip())
|
100 |
-
cleaned_response = '. '.join(cleaned_sentences).strip()
|
101 |
-
if cleaned_response and not cleaned_response.endswith((".", "!", "?")):
|
102 |
-
cleaned_response += "."
|
103 |
-
return cleaned_response
|
104 |
-
|
105 |
-
# Gradio interface and application logic
|
106 |
-
def query_model(question):
|
107 |
-
"""
|
108 |
-
Process a question through the model and return the response.
|
109 |
-
|
110 |
-
Args:
|
111 |
-
question (str): The question submitted by the user.
|
112 |
-
|
113 |
-
Returns:
|
114 |
-
str: Generated response or welcome message if no question is provided.
|
115 |
-
"""
|
116 |
-
if question == "":
|
117 |
-
return welcome_message
|
118 |
-
relevant_segment = find_relevant_segment(question, segments)
|
119 |
-
response = generate_response(question, relevant_segment)
|
120 |
-
return response
|
121 |
-
|
122 |
-
with gr.Blocks() as demo:
|
123 |
-
gr.Markdown(welcome_message)
|
124 |
-
with gr.Row():
|
125 |
-
with gr.Column():
|
126 |
-
gr.Markdown(topics)
|
127 |
-
with gr.Column():
|
128 |
-
gr.Markdown(countries)
|
129 |
-
with gr.Row():
|
130 |
-
img = gr.Image(os.path.join(os.getcwd(), "final.png"), width=500)
|
131 |
-
with gr.Row():
|
132 |
-
with gr.Column():
|
133 |
-
question = gr.Textbox(label="Your question", placeholder="What do you want to ask about?")
|
134 |
-
answer = gr.Textbox(label="VisaBot Response", placeholder="VisaBot will respond here...", interactive=False, lines=10)
|
135 |
-
submit_button = gr.Button("Submit")
|
136 |
-
submit_button.click(fn=query_model, inputs=question, outputs=answer)
|
137 |
-
|
138 |
-
demo.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|