Spaces:
Sleeping
Sleeping
Delete app8.py
Browse files
app8.py
DELETED
@@ -1,409 +0,0 @@
|
|
1 |
-
import streamlit as st
|
2 |
-
import pandas as pd
|
3 |
-
import os
|
4 |
-
from datetime import datetime
|
5 |
-
import random
|
6 |
-
from pathlib import Path
|
7 |
-
from openai import OpenAI
|
8 |
-
from dotenv import load_dotenv
|
9 |
-
from langchain_core.prompts import PromptTemplate
|
10 |
-
|
11 |
-
# Load environment variables
|
12 |
-
load_dotenv()
|
13 |
-
##openai_api_key = os.getenv("OPENAI_API_KEY")
|
14 |
-
|
15 |
-
# Initialize the client
|
16 |
-
client = OpenAI(
|
17 |
-
base_url="https://api-inference.huggingface.co/v1",
|
18 |
-
api_key=os.environ.get('TOKEN2') # Add your Huggingface token here
|
19 |
-
)
|
20 |
-
|
21 |
-
|
22 |
-
# Initialize OpenAI client
|
23 |
-
##client = OpenAI(api_key=os.getenv('OPENAI_API_KEY'))
|
24 |
-
|
25 |
-
# Custom CSS for better appearance
|
26 |
-
st.markdown("""
|
27 |
-
<style>
|
28 |
-
.stButton > button {
|
29 |
-
width: 100%;
|
30 |
-
margin-bottom: 10px;
|
31 |
-
background-color: #4CAF50;
|
32 |
-
color: white;
|
33 |
-
border: none;
|
34 |
-
padding: 10px;
|
35 |
-
border-radius: 5px;
|
36 |
-
}
|
37 |
-
.task-button {
|
38 |
-
background-color: #2196F3 !important;
|
39 |
-
}
|
40 |
-
.stSelectbox {
|
41 |
-
margin-bottom: 20px;
|
42 |
-
}
|
43 |
-
.output-container {
|
44 |
-
padding: 20px;
|
45 |
-
border-radius: 5px;
|
46 |
-
border: 1px solid #ddd;
|
47 |
-
margin: 10px 0;
|
48 |
-
}
|
49 |
-
.status-container {
|
50 |
-
padding: 10px;
|
51 |
-
border-radius: 5px;
|
52 |
-
margin: 10px 0;
|
53 |
-
}
|
54 |
-
.sidebar-info {
|
55 |
-
padding: 10px;
|
56 |
-
background-color: #f0f2f6;
|
57 |
-
border-radius: 5px;
|
58 |
-
margin: 10px 0;
|
59 |
-
}
|
60 |
-
</style>
|
61 |
-
""", unsafe_allow_html=True)
|
62 |
-
|
63 |
-
# Create data directories if they don't exist
|
64 |
-
if not os.path.exists('data'):
|
65 |
-
os.makedirs('data')
|
66 |
-
|
67 |
-
def read_csv_with_encoding(file):
|
68 |
-
encodings = ['utf-8', 'latin1', 'iso-8859-1', 'cp1252']
|
69 |
-
for encoding in encodings:
|
70 |
-
try:
|
71 |
-
return pd.read_csv(file, encoding=encoding)
|
72 |
-
except UnicodeDecodeError:
|
73 |
-
continue
|
74 |
-
raise UnicodeDecodeError("Failed to read file with any supported encoding")
|
75 |
-
|
76 |
-
def save_to_csv(data, filename):
|
77 |
-
df = pd.DataFrame(data)
|
78 |
-
df.to_csv(f'data/{filename}', index=False)
|
79 |
-
return df
|
80 |
-
|
81 |
-
def load_from_csv(filename):
|
82 |
-
try:
|
83 |
-
return pd.read_csv(f'data/{filename}')
|
84 |
-
except:
|
85 |
-
return pd.DataFrame()
|
86 |
-
|
87 |
-
# Define reset function
|
88 |
-
def reset_conversation():
|
89 |
-
st.session_state.conversation = []
|
90 |
-
st.session_state.messages = []
|
91 |
-
|
92 |
-
# Initialize session state
|
93 |
-
if "messages" not in st.session_state:
|
94 |
-
st.session_state.messages = []
|
95 |
-
|
96 |
-
# Main app title
|
97 |
-
st.title("🤖 Text Data Generation & Labeling App")
|
98 |
-
|
99 |
-
# Sidebar settings
|
100 |
-
with st.sidebar:
|
101 |
-
st.title("⚙️ Settings")
|
102 |
-
|
103 |
-
selected_model = st.selectbox(
|
104 |
-
"Select Model",
|
105 |
-
["meta-llama/Meta-Llama-3-8B-Instruct"],
|
106 |
-
key='model_select'
|
107 |
-
)
|
108 |
-
|
109 |
-
temperature = st.slider(
|
110 |
-
"Temperature",
|
111 |
-
0.0, 1.0, 0.5,
|
112 |
-
help="Controls randomness in generation"
|
113 |
-
)
|
114 |
-
|
115 |
-
st.button("🔄 Reset Conversation", on_click=reset_conversation)
|
116 |
-
|
117 |
-
with st.container():
|
118 |
-
st.markdown("""
|
119 |
-
<div class="sidebar-info">
|
120 |
-
<h4>Current Model: {}</h4>
|
121 |
-
<p><em>Note: Generated content may be inaccurate or false.</em></p>
|
122 |
-
</div>
|
123 |
-
""".format(selected_model), unsafe_allow_html=True)
|
124 |
-
|
125 |
-
# Main content
|
126 |
-
col1, col2 = st.columns(2)
|
127 |
-
|
128 |
-
with col1:
|
129 |
-
if st.button("📝 Data Generation", key="gen_button", help="Generate new data"):
|
130 |
-
st.session_state.task_choice = "Data Generation"
|
131 |
-
|
132 |
-
with col2:
|
133 |
-
if st.button("🏷️ Data Labeling", key="label_button", help="Label existing data"):
|
134 |
-
st.session_state.task_choice = "Data Labeling"
|
135 |
-
|
136 |
-
if "task_choice" in st.session_state:
|
137 |
-
if st.session_state.task_choice == "Data Generation":
|
138 |
-
st.header("📝 Data Generation")
|
139 |
-
|
140 |
-
classification_type = st.selectbox(
|
141 |
-
"Classification Type",
|
142 |
-
["Sentiment Analysis", "Binary Classification", "Multi-Class Classification"]
|
143 |
-
)
|
144 |
-
|
145 |
-
if classification_type == "Sentiment Analysis":
|
146 |
-
labels = ["Positive", "Negative", "Neutral"]
|
147 |
-
elif classification_type == "Binary Classification":
|
148 |
-
col1, col2 = st.columns(2)
|
149 |
-
with col1:
|
150 |
-
label_1 = st.text_input("First class", "Positive")
|
151 |
-
with col2:
|
152 |
-
label_2 = st.text_input("Second class", "Negative")
|
153 |
-
labels = [label_1, label_2] if label_1 and label_2 else ["Positive", "Negative"]
|
154 |
-
else:
|
155 |
-
num_classes = st.slider("Number of classes", 3, 10, 3)
|
156 |
-
labels = []
|
157 |
-
cols = st.columns(3)
|
158 |
-
for i in range(num_classes):
|
159 |
-
with cols[i % 3]:
|
160 |
-
label = st.text_input(f"Class {i+1}", f"Class_{i+1}")
|
161 |
-
labels.append(label)
|
162 |
-
|
163 |
-
domain = st.selectbox("Domain", ["Restaurant reviews", "E-commerce reviews", "Custom"])
|
164 |
-
if domain == "Custom":
|
165 |
-
domain = st.text_input("Specify custom domain")
|
166 |
-
|
167 |
-
col1, col2 = st.columns(2)
|
168 |
-
with col1:
|
169 |
-
min_words = st.number_input("Min words", 10, 90, 20)
|
170 |
-
with col2:
|
171 |
-
max_words = st.number_input("Max words", min_words, 90, 50)
|
172 |
-
|
173 |
-
use_few_shot = st.toggle("Use few-shot examples")
|
174 |
-
few_shot_examples = []
|
175 |
-
if use_few_shot:
|
176 |
-
num_examples = st.slider("Number of few-shot examples", 1, 5, 1)
|
177 |
-
for i in range(num_examples):
|
178 |
-
with st.expander(f"Example {i+1}"):
|
179 |
-
content = st.text_area(f"Content", key=f"few_shot_content_{i}")
|
180 |
-
label = st.selectbox(f"Label", labels, key=f"few_shot_label_{i}")
|
181 |
-
if content and label:
|
182 |
-
few_shot_examples.append({"content": content, "label": label})
|
183 |
-
|
184 |
-
num_to_generate = st.number_input("Number of examples", 1, 100, 10)
|
185 |
-
user_prompt = st.text_area("Additional instructions (optional)")
|
186 |
-
|
187 |
-
# Updated prompt template with word length constraints
|
188 |
-
prompt_template = PromptTemplate(
|
189 |
-
input_variables=["classification_type", "domain", "num_examples", "min_words", "max_words", "labels", "user_prompt"],
|
190 |
-
template=(
|
191 |
-
"You are a professional {classification_type} expert tasked with generating examples for {domain}.\n"
|
192 |
-
"Use the following parameters:\n"
|
193 |
-
"- Generate exactly {num_examples} examples\n"
|
194 |
-
"- Each example MUST be between {min_words} and {max_words} words long\n"
|
195 |
-
"- Use these labels: {labels}\n"
|
196 |
-
"- Generate the examples in this format: 'Example text. Label: [label]'\n"
|
197 |
-
"- Do not include word counts or any additional information\n"
|
198 |
-
"Additional instructions: {user_prompt}\n\n"
|
199 |
-
"Generate numbered examples:"
|
200 |
-
)
|
201 |
-
)
|
202 |
-
|
203 |
-
col1, col2 = st.columns(2)
|
204 |
-
with col1:
|
205 |
-
if st.button("🎯 Generate Examples"):
|
206 |
-
with st.spinner("Generating examples..."):
|
207 |
-
system_prompt = prompt_template.format(
|
208 |
-
classification_type=classification_type,
|
209 |
-
domain=domain,
|
210 |
-
num_examples=num_to_generate,
|
211 |
-
min_words=min_words,
|
212 |
-
max_words=max_words,
|
213 |
-
labels=", ".join(labels),
|
214 |
-
user_prompt=user_prompt
|
215 |
-
)
|
216 |
-
try:
|
217 |
-
stream = client.chat.completions.create(
|
218 |
-
model=selected_model,
|
219 |
-
messages=[{"role": "system", "content": system_prompt}],
|
220 |
-
temperature=temperature,
|
221 |
-
stream=True,
|
222 |
-
max_tokens=3000,
|
223 |
-
)
|
224 |
-
response = st.write_stream(stream)
|
225 |
-
st.session_state.messages.append({"role": "assistant", "content": response})
|
226 |
-
except Exception as e:
|
227 |
-
st.error("An error occurred during generation.")
|
228 |
-
st.error(f"Details: {e}")
|
229 |
-
|
230 |
-
with col2:
|
231 |
-
if st.button("🔄 Regenerate"):
|
232 |
-
st.session_state.messages = st.session_state.messages[:-1] if st.session_state.messages else []
|
233 |
-
with st.spinner("Regenerating examples..."):
|
234 |
-
system_prompt = prompt_template.format(
|
235 |
-
classification_type=classification_type,
|
236 |
-
domain=domain,
|
237 |
-
num_examples=num_to_generate,
|
238 |
-
min_words=min_words,
|
239 |
-
max_words=max_words,
|
240 |
-
labels=", ".join(labels),
|
241 |
-
user_prompt=user_prompt
|
242 |
-
)
|
243 |
-
try:
|
244 |
-
stream = client.chat.completions.create(
|
245 |
-
model=selected_model,
|
246 |
-
messages=[{"role": "system", "content": system_prompt}],
|
247 |
-
temperature=temperature,
|
248 |
-
stream=True,
|
249 |
-
max_tokens=3000,
|
250 |
-
)
|
251 |
-
response = st.write_stream(stream)
|
252 |
-
st.session_state.messages.append({"role": "assistant", "content": response})
|
253 |
-
except Exception as e:
|
254 |
-
st.error("An error occurred during regeneration.")
|
255 |
-
st.error(f"Details: {e}")
|
256 |
-
|
257 |
-
elif st.session_state.task_choice == "Data Labeling":
|
258 |
-
st.header("🏷️ Data Labeling")
|
259 |
-
|
260 |
-
classification_type = st.selectbox(
|
261 |
-
"Classification Type",
|
262 |
-
["Sentiment Analysis", "Binary Classification", "Multi-Class Classification"],
|
263 |
-
key="label_class_type"
|
264 |
-
)
|
265 |
-
|
266 |
-
if classification_type == "Sentiment Analysis":
|
267 |
-
labels = ["Positive", "Negative", "Neutral"]
|
268 |
-
elif classification_type == "Binary Classification":
|
269 |
-
col1, col2 = st.columns(2)
|
270 |
-
with col1:
|
271 |
-
label_1 = st.text_input("First class", "Positive", key="label_first")
|
272 |
-
with col2:
|
273 |
-
label_2 = st.text_input("Second class", "Negative", key="label_second")
|
274 |
-
labels = [label_1, label_2] if label_1 and label_2 else ["Positive", "Negative"]
|
275 |
-
else:
|
276 |
-
num_classes = st.slider("Number of classes", 3, 10, 3, key="label_num_classes")
|
277 |
-
labels = []
|
278 |
-
cols = st.columns(3)
|
279 |
-
for i in range(num_classes):
|
280 |
-
with cols[i % 3]:
|
281 |
-
label = st.text_input(f"Class {i+1}", f"Class_{i+1}", key=f"label_class_{i}")
|
282 |
-
labels.append(label)
|
283 |
-
|
284 |
-
use_few_shot = st.toggle("Use few-shot examples for labeling")
|
285 |
-
few_shot_examples = []
|
286 |
-
if use_few_shot:
|
287 |
-
num_few_shot = st.slider("Number of few-shot examples", 1, 5, 1)
|
288 |
-
for i in range(num_few_shot):
|
289 |
-
with st.expander(f"Few-shot Example {i+1}"):
|
290 |
-
content = st.text_area(f"Content", key=f"label_few_shot_content_{i}")
|
291 |
-
label = st.selectbox(f"Label", labels, key=f"label_few_shot_label_{i}")
|
292 |
-
if content and label:
|
293 |
-
few_shot_examples.append(f"{content}\nLabel: {label}")
|
294 |
-
|
295 |
-
num_examples = st.number_input("Number of examples to classify", 1, 100, 1)
|
296 |
-
|
297 |
-
examples_to_classify = []
|
298 |
-
if num_examples <= 20:
|
299 |
-
for i in range(num_examples):
|
300 |
-
example = st.text_area(f"Example {i+1}", key=f"example_{i}")
|
301 |
-
if example:
|
302 |
-
examples_to_classify.append(example)
|
303 |
-
else:
|
304 |
-
examples_text = st.text_area(
|
305 |
-
"Enter examples (one per line)",
|
306 |
-
height=300,
|
307 |
-
help="Enter each example on a new line"
|
308 |
-
)
|
309 |
-
if examples_text:
|
310 |
-
examples_to_classify = [ex.strip() for ex in examples_text.split('\n') if ex.strip()]
|
311 |
-
if len(examples_to_classify) > num_examples:
|
312 |
-
examples_to_classify = examples_to_classify[:num_examples]
|
313 |
-
|
314 |
-
user_prompt = st.text_area("Additional instructions (optional)", key="label_instructions")
|
315 |
-
|
316 |
-
# Updated prompt template for labeling
|
317 |
-
few_shot_text = "\n\n".join(few_shot_examples) if few_shot_examples else ""
|
318 |
-
examples_text = "\n".join(f"{i+1}. {ex}" for i, ex in enumerate(examples_to_classify))
|
319 |
-
|
320 |
-
|
321 |
-
label_prompt_template = PromptTemplate(
|
322 |
-
input_variables=["classification_type", "labels", "few_shot_examples", "examples", "user_prompt"],
|
323 |
-
template=(
|
324 |
-
"You are a professional {classification_type} expert. Classify the following examples using these labels: {labels}.\n"
|
325 |
-
"Instructions:\n"
|
326 |
-
"- Return ONLY the numbered example followed by its classification\n"
|
327 |
-
"- Use the format: 'Example text. Label: [label]'\n"
|
328 |
-
"- Do not provide explanations or justifications\n"
|
329 |
-
"{user_prompt}\n\n"
|
330 |
-
"Few-shot examples:\n{few_shot_examples}\n\n"
|
331 |
-
"Examples to classify:\n{examples}\n\n"
|
332 |
-
"Output:\n"
|
333 |
-
)
|
334 |
-
)
|
335 |
-
col1, col2 = st.columns(2)
|
336 |
-
with col1:
|
337 |
-
if st.button("🏷️ Label Data"):
|
338 |
-
if examples_to_classify:
|
339 |
-
with st.spinner("Labeling data..."):
|
340 |
-
system_prompt = label_prompt_template.format(
|
341 |
-
classification_type=classification_type,
|
342 |
-
labels=", ".join(labels),
|
343 |
-
few_shot_examples=few_shot_text,
|
344 |
-
examples=examples_text,
|
345 |
-
user_prompt=user_prompt
|
346 |
-
)
|
347 |
-
try:
|
348 |
-
stream = client.chat.completions.create(
|
349 |
-
model=selected_model,
|
350 |
-
messages=[{"role": "system", "content": system_prompt}],
|
351 |
-
temperature=temperature,
|
352 |
-
stream=True,
|
353 |
-
max_tokens=3000,
|
354 |
-
)
|
355 |
-
response = st.write_stream(stream)
|
356 |
-
st.session_state.messages.append({"role": "assistant", "content": response})
|
357 |
-
except Exception as e:
|
358 |
-
st.error("An error occurred during labeling.")
|
359 |
-
st.error(f"Details: {e}")
|
360 |
-
else:
|
361 |
-
st.warning("Please enter at least one example to classify.")
|
362 |
-
|
363 |
-
with col2:
|
364 |
-
if st.button("🔄 Relabel"):
|
365 |
-
if examples_to_classify:
|
366 |
-
st.session_state.messages = st.session_state.messages[:-1] if st.session_state.messages else []
|
367 |
-
with st.spinner("Relabeling data..."):
|
368 |
-
system_prompt = label_prompt_template.format(
|
369 |
-
classification_type=classification_type,
|
370 |
-
labels=", ".join(labels),
|
371 |
-
few_shot_examples=few_shot_text,
|
372 |
-
examples=examples_text,
|
373 |
-
user_prompt=user_prompt
|
374 |
-
)
|
375 |
-
try:
|
376 |
-
stream = client.chat.completions.create(
|
377 |
-
model=selected_model,
|
378 |
-
messages=[{"role": "system", "content": system_prompt}],
|
379 |
-
temperature=temperature,
|
380 |
-
stream=True,
|
381 |
-
max_tokens=3000,
|
382 |
-
)
|
383 |
-
response = st.write_stream(stream)
|
384 |
-
st.session_state.messages.append({"role": "assistant", "content": response})
|
385 |
-
except Exception as e:
|
386 |
-
st.error("An error occurred during relabeling.")
|
387 |
-
st.error(f"Details: {e}")
|
388 |
-
else:
|
389 |
-
st.warning("Please enter at least one example to classify.")
|
390 |
-
|
391 |
-
if st.session_state.messages:
|
392 |
-
st.markdown("### Output:")
|
393 |
-
for message in st.session_state.messages[-1:]:
|
394 |
-
st.markdown(message["content"])
|
395 |
-
|
396 |
-
##if st.session_state.messages:
|
397 |
-
##st.markdown("### Output:")
|
398 |
-
##last_message = st.session_state.messages[-1]["content"]
|
399 |
-
|
400 |
-
# Find the position of "Output:" if it exists
|
401 |
-
##output_start = last_message.find("Output:")
|
402 |
-
|
403 |
-
##if output_start != -1:
|
404 |
-
# Display only the content after "Output:"
|
405 |
-
##cleaned_output = last_message[output_start + 7:].strip()
|
406 |
-
##st.markdown(cleaned_output)
|
407 |
-
##else:
|
408 |
-
# If "Output:" is not found, display the content as is
|
409 |
-
##st.markdown(last_message)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|