Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -2,13 +2,11 @@
|
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
import requests
|
5 |
-
import re
|
6 |
-
from datetime import datetime
|
7 |
import json
|
8 |
-
|
9 |
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
|
10 |
from sentence_transformers import SentenceTransformer
|
11 |
from qdrant_client import QdrantClient
|
|
|
12 |
|
13 |
# === Load Models ===
|
14 |
print("Loading zero-shot classifier...")
|
@@ -17,43 +15,32 @@ classifier = pipeline("zero-shot-classification", model="facebook/bart-large-mnl
|
|
17 |
print("Loading embedding model...")
|
18 |
embedding_model = SentenceTransformer("intfloat/e5-large")
|
19 |
|
20 |
-
print("Loading
|
21 |
-
tokenizer = AutoTokenizer.from_pretrained("
|
22 |
-
model = AutoModelForCausalLM.from_pretrained(
|
23 |
-
"WizardLM/WizardMath-7B-V1.1", torch_dtype=torch.float16, device_map="auto"
|
24 |
-
)
|
25 |
|
26 |
# === Qdrant Setup ===
|
27 |
print("Connecting to Qdrant...")
|
28 |
qdrant_client = QdrantClient(path="qdrant_data")
|
29 |
collection_name = "math_problems"
|
30 |
|
31 |
-
# === Guard
|
32 |
def is_valid_math_question(text):
|
33 |
candidate_labels = ["math", "not math"]
|
34 |
result = classifier(text, candidate_labels)
|
35 |
return result['labels'][0] == "math" and result['scores'][0] > 0.7
|
36 |
|
37 |
-
|
38 |
-
if not answer or len(answer.strip()) < 10:
|
39 |
-
return False
|
40 |
-
math_keywords = ["solve", "equation", "integral", "derivative", "value", "expression", "steps", "solution"]
|
41 |
-
if not any(word in answer.lower() for word in math_keywords):
|
42 |
-
return False
|
43 |
-
banned_keywords = ["kill", "bomb", "hate", "politics", "violence"]
|
44 |
-
if any(word in answer.lower() for word in banned_keywords):
|
45 |
-
return False
|
46 |
-
if re.match(r"^\s*I'm just a model|Sorry, I can't|As an AI", answer, re.IGNORECASE):
|
47 |
-
return False
|
48 |
-
return True
|
49 |
-
|
50 |
-
# === Retrieval ===
|
51 |
def retrieve_from_qdrant(query):
|
52 |
query_vector = embedding_model.encode(query).tolist()
|
53 |
-
hits = qdrant_client.
|
|
|
|
|
|
|
|
|
54 |
return [hit.payload for hit in hits] if hits else []
|
55 |
|
56 |
-
# === Web Search ===
|
57 |
def web_search_tavily(query):
|
58 |
TAVILY_API_KEY = "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"
|
59 |
response = requests.post(
|
@@ -62,14 +49,10 @@ def web_search_tavily(query):
|
|
62 |
)
|
63 |
return response.json().get("answer", "No answer found from Tavily.")
|
64 |
|
65 |
-
# ===
|
66 |
-
def generate_step_by_step_answer(question, context
|
67 |
-
prompt = f"
|
68 |
-
|
69 |
-
prompt += f"### Context:\n{context}\n"
|
70 |
-
prompt += "### Let's solve it step by step:\n"
|
71 |
-
|
72 |
-
inputs = tokenizer(prompt, return_tensors="pt").to("cuda")
|
73 |
outputs = model.generate(
|
74 |
**inputs,
|
75 |
max_new_tokens=256,
|
@@ -78,67 +61,65 @@ def generate_step_by_step_answer(question, context=""):
|
|
78 |
do_sample=True,
|
79 |
pad_token_id=tokenizer.eos_token_id
|
80 |
)
|
81 |
-
|
82 |
-
answer = decoded.split("### Let's solve it step by step:")[-1].strip()
|
83 |
-
return answer
|
84 |
|
85 |
# === Router ===
|
86 |
def router(question):
|
87 |
if not is_valid_math_question(question):
|
88 |
-
return "โ Only math questions are accepted. Please rephrase."
|
89 |
-
|
90 |
-
context_items = retrieve_from_qdrant(question)
|
91 |
-
context = "\n".join([item.get("solution", "") for item in context_items])
|
92 |
|
|
|
|
|
93 |
if context:
|
94 |
answer = generate_step_by_step_answer(question, context)
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
99 |
-
return answer if output_guardrails(answer) else "โ ๏ธ No valid math answer found."
|
100 |
|
101 |
# === Feedback Storage ===
|
102 |
-
def store_feedback(question, answer,
|
103 |
entry = {
|
104 |
"question": question,
|
105 |
"model_answer": answer,
|
106 |
-
"feedback": feedback,
|
107 |
"correct_answer": correct_answer,
|
108 |
"timestamp": str(datetime.now())
|
109 |
}
|
110 |
with open("feedback.json", "a") as f:
|
111 |
f.write(json.dumps(entry) + "\n")
|
112 |
|
113 |
-
# === Gradio
|
114 |
def ask_question(question):
|
115 |
-
answer = router(question)
|
116 |
return answer, question, answer
|
117 |
|
118 |
-
def submit_feedback(question, model_answer,
|
119 |
-
store_feedback(question, model_answer,
|
120 |
return "โ
Feedback received. Thank you!"
|
121 |
|
|
|
122 |
with gr.Blocks() as demo:
|
123 |
-
gr.Markdown("## ๐งฎ Math
|
124 |
|
125 |
with gr.Row():
|
126 |
question_input = gr.Textbox(label="Enter your math question", lines=2)
|
127 |
submit_btn = gr.Button("Get Answer")
|
128 |
|
129 |
-
answer_output = gr.Markdown()
|
130 |
hidden_q = gr.Textbox(visible=False)
|
131 |
hidden_a = gr.Textbox(visible=False)
|
132 |
|
133 |
submit_btn.click(fn=ask_question, inputs=[question_input], outputs=[answer_output, hidden_q, hidden_a])
|
134 |
|
135 |
-
gr.Markdown("### ๐ Feedback")
|
136 |
-
|
137 |
-
|
138 |
fb_status = gr.Textbox(label="Status", interactive=False)
|
139 |
|
140 |
-
|
141 |
-
|
142 |
-
|
|
|
|
|
143 |
|
144 |
-
demo.launch(share=True
|
|
|
2 |
import gradio as gr
|
3 |
import torch
|
4 |
import requests
|
|
|
|
|
5 |
import json
|
|
|
6 |
from transformers import pipeline, AutoTokenizer, AutoModelForCausalLM
|
7 |
from sentence_transformers import SentenceTransformer
|
8 |
from qdrant_client import QdrantClient
|
9 |
+
from datetime import datetime
|
10 |
|
11 |
# === Load Models ===
|
12 |
print("Loading zero-shot classifier...")
|
|
|
15 |
print("Loading embedding model...")
|
16 |
embedding_model = SentenceTransformer("intfloat/e5-large")
|
17 |
|
18 |
+
print("Loading step-by-step generator...")
|
19 |
+
tokenizer = AutoTokenizer.from_pretrained("microsoft/phi-2")
|
20 |
+
model = AutoModelForCausalLM.from_pretrained("microsoft/phi-2")
|
|
|
|
|
21 |
|
22 |
# === Qdrant Setup ===
|
23 |
print("Connecting to Qdrant...")
|
24 |
qdrant_client = QdrantClient(path="qdrant_data")
|
25 |
collection_name = "math_problems"
|
26 |
|
27 |
+
# === Guard Function ===
|
28 |
def is_valid_math_question(text):
|
29 |
candidate_labels = ["math", "not math"]
|
30 |
result = classifier(text, candidate_labels)
|
31 |
return result['labels'][0] == "math" and result['scores'][0] > 0.7
|
32 |
|
33 |
+
# === Retrieval from Qdrant ===
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
34 |
def retrieve_from_qdrant(query):
|
35 |
query_vector = embedding_model.encode(query).tolist()
|
36 |
+
hits = qdrant_client.query_points(
|
37 |
+
collection_name=collection_name,
|
38 |
+
query_vector=query_vector,
|
39 |
+
limit=3
|
40 |
+
)
|
41 |
return [hit.payload for hit in hits] if hits else []
|
42 |
|
43 |
+
# === Web Search Fallback ===
|
44 |
def web_search_tavily(query):
|
45 |
TAVILY_API_KEY = "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"
|
46 |
response = requests.post(
|
|
|
49 |
)
|
50 |
return response.json().get("answer", "No answer found from Tavily.")
|
51 |
|
52 |
+
# === Generator ===
|
53 |
+
def generate_step_by_step_answer(question, context):
|
54 |
+
prompt = f"Answer the following math question step-by-step:\nQuestion: {question}\nContext: {context}\nAnswer:"
|
55 |
+
inputs = tokenizer(prompt, return_tensors="pt")
|
|
|
|
|
|
|
|
|
56 |
outputs = model.generate(
|
57 |
**inputs,
|
58 |
max_new_tokens=256,
|
|
|
61 |
do_sample=True,
|
62 |
pad_token_id=tokenizer.eos_token_id
|
63 |
)
|
64 |
+
return tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
|
|
65 |
|
66 |
# === Router ===
|
67 |
def router(question):
|
68 |
if not is_valid_math_question(question):
|
69 |
+
return "โ Only math questions are accepted. Please rephrase.", ""
|
|
|
|
|
|
|
70 |
|
71 |
+
retrieved = retrieve_from_qdrant(question)
|
72 |
+
context = "\n".join([item["solution"] for item in retrieved if "solution" in item])
|
73 |
if context:
|
74 |
answer = generate_step_by_step_answer(question, context)
|
75 |
+
return answer, context
|
76 |
+
else:
|
77 |
+
fallback = web_search_tavily(question)
|
78 |
+
return fallback, "Tavily Search"
|
|
|
79 |
|
80 |
# === Feedback Storage ===
|
81 |
+
def store_feedback(question, answer, correct_answer):
|
82 |
entry = {
|
83 |
"question": question,
|
84 |
"model_answer": answer,
|
|
|
85 |
"correct_answer": correct_answer,
|
86 |
"timestamp": str(datetime.now())
|
87 |
}
|
88 |
with open("feedback.json", "a") as f:
|
89 |
f.write(json.dumps(entry) + "\n")
|
90 |
|
91 |
+
# === Gradio Functions ===
|
92 |
def ask_question(question):
|
93 |
+
answer, context = router(question)
|
94 |
return answer, question, answer
|
95 |
|
96 |
+
def submit_feedback(question, model_answer, correct_answer):
|
97 |
+
store_feedback(question, model_answer, correct_answer)
|
98 |
return "โ
Feedback received. Thank you!"
|
99 |
|
100 |
+
# === Gradio UI ===
|
101 |
with gr.Blocks() as demo:
|
102 |
+
gr.Markdown("## ๐งฎ Math Question Answering with Retrieval + Feedback")
|
103 |
|
104 |
with gr.Row():
|
105 |
question_input = gr.Textbox(label="Enter your math question", lines=2)
|
106 |
submit_btn = gr.Button("Get Answer")
|
107 |
|
108 |
+
answer_output = gr.Markdown(label="Answer")
|
109 |
hidden_q = gr.Textbox(visible=False)
|
110 |
hidden_a = gr.Textbox(visible=False)
|
111 |
|
112 |
submit_btn.click(fn=ask_question, inputs=[question_input], outputs=[answer_output, hidden_q, hidden_a])
|
113 |
|
114 |
+
gr.Markdown("### ๐ Submit Feedback")
|
115 |
+
fb_correct = gr.Textbox(label="Correct Answer (optional)")
|
116 |
+
fb_submit = gr.Button("Submit Feedback")
|
117 |
fb_status = gr.Textbox(label="Status", interactive=False)
|
118 |
|
119 |
+
fb_submit.click(
|
120 |
+
fn=submit_feedback,
|
121 |
+
inputs=[hidden_q, hidden_a, fb_correct],
|
122 |
+
outputs=[fb_status]
|
123 |
+
)
|
124 |
|
125 |
+
demo.launch(share=True)
|