Spaces:
Sleeping
Sleeping
Commit
·
a27324e
1
Parent(s):
3bf5e4f
updated app.py
Browse files
app.py
CHANGED
@@ -5,11 +5,17 @@ from peft import PeftModel
|
|
5 |
import time
|
6 |
import gc
|
7 |
import os
|
|
|
8 |
|
9 |
# Configuration
|
10 |
BASE_MODEL = "microsoft/phi-2"
|
11 |
ADAPTER_MODEL = "pradeep6kumar2024/phi2-qlora-assistant"
|
12 |
|
|
|
|
|
|
|
|
|
|
|
13 |
class ModelWrapper:
|
14 |
def __init__(self):
|
15 |
self.model = None
|
@@ -26,6 +32,8 @@ class ModelWrapper:
|
|
26 |
# Clear memory
|
27 |
gc.collect()
|
28 |
|
|
|
|
|
29 |
print("Loading tokenizer...")
|
30 |
self.tokenizer = AutoTokenizer.from_pretrained(
|
31 |
BASE_MODEL,
|
@@ -34,21 +42,26 @@ class ModelWrapper:
|
|
34 |
)
|
35 |
self.tokenizer.pad_token = self.tokenizer.eos_token
|
36 |
|
|
|
|
|
37 |
print("Loading base model...")
|
38 |
base_model = AutoModelForCausalLM.from_pretrained(
|
39 |
BASE_MODEL,
|
40 |
-
torch_dtype=torch.float32,
|
41 |
device_map="cpu",
|
42 |
trust_remote_code=True,
|
43 |
use_flash_attention_2=False,
|
44 |
-
low_cpu_mem_usage=True
|
|
|
45 |
)
|
46 |
|
|
|
|
|
47 |
print("Loading LoRA adapter...")
|
48 |
self.model = PeftModel.from_pretrained(
|
49 |
base_model,
|
50 |
ADAPTER_MODEL,
|
51 |
-
torch_dtype=torch.float32,
|
52 |
device_map="cpu"
|
53 |
)
|
54 |
|
@@ -56,6 +69,8 @@ class ModelWrapper:
|
|
56 |
del base_model
|
57 |
gc.collect()
|
58 |
|
|
|
|
|
59 |
self.model.eval()
|
60 |
print("Model loading complete!")
|
61 |
self.loaded = True
|
@@ -63,188 +78,79 @@ class ModelWrapper:
|
|
63 |
print(f"Error during model loading: {str(e)}")
|
64 |
raise
|
65 |
|
66 |
-
def generate_response(self, prompt, max_length=
|
67 |
if not self.loaded:
|
68 |
self.load_model()
|
69 |
|
70 |
try:
|
71 |
-
#
|
72 |
if "function" in prompt.lower() and "python" in prompt.lower():
|
73 |
-
enhanced_prompt = f"""Write
|
74 |
-
{prompt}
|
75 |
-
Include:
|
76 |
-
- Function implementation with comments
|
77 |
-
- Example usage
|
78 |
-
- Output demonstration
|
79 |
-
|
80 |
-
Provide only the implementation, no conversation."""
|
81 |
elif any(word in prompt.lower() for word in ["explain", "what is", "how does", "describe"]):
|
82 |
-
enhanced_prompt = f"""
|
83 |
-
|
84 |
-
{prompt}
|
85 |
-
|
86 |
-
Your response should include:
|
87 |
-
1. A clear explanation in simple terms
|
88 |
-
2. Practical examples and applications
|
89 |
-
3. Important concepts to understand
|
90 |
-
|
91 |
-
End your response when the explanation is complete. Do not ask questions or engage in conversation."""
|
92 |
else:
|
93 |
-
enhanced_prompt =
|
94 |
-
|
95 |
-
{prompt}
|
96 |
-
|
97 |
-
End your response when complete. Do not ask questions or engage in conversation."""
|
98 |
|
99 |
-
print(f"Enhanced prompt: {enhanced_prompt}")
|
100 |
|
101 |
-
# Tokenize input
|
102 |
inputs = self.tokenizer(
|
103 |
enhanced_prompt,
|
104 |
return_tensors="pt",
|
105 |
truncation=True,
|
106 |
-
max_length=
|
107 |
padding=True
|
108 |
-
).to("cpu")
|
109 |
|
110 |
-
# Generate with
|
111 |
start_time = time.time()
|
112 |
with torch.no_grad():
|
113 |
outputs = self.model.generate(
|
114 |
**inputs,
|
115 |
-
max_length=min(max_length,
|
116 |
-
min_length=
|
117 |
temperature=min(0.5, temperature),
|
118 |
top_p=min(0.85, top_p),
|
119 |
do_sample=True,
|
120 |
pad_token_id=self.tokenizer.pad_token_id,
|
121 |
eos_token_id=self.tokenizer.eos_token_id,
|
122 |
-
repetition_penalty=1.
|
123 |
-
no_repeat_ngram_size=
|
124 |
num_return_sequences=1,
|
125 |
early_stopping=True,
|
126 |
-
num_beams=
|
127 |
length_penalty=0.6
|
128 |
)
|
129 |
|
130 |
# Decode response
|
131 |
response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
132 |
-
print(f"Raw response: {response}") # Debug logging
|
133 |
|
134 |
# Clean up the response
|
135 |
if response.startswith(enhanced_prompt):
|
136 |
response = response[len(enhanced_prompt):].strip()
|
137 |
|
138 |
-
|
139 |
-
|
140 |
-
# Remove common closure patterns and conversation starters
|
141 |
-
closures = [
|
142 |
-
"Best regards,",
|
143 |
-
"Sincerely,",
|
144 |
-
"Thanks,",
|
145 |
-
"Thank you,",
|
146 |
-
"Regards,",
|
147 |
-
"Assistant:",
|
148 |
-
"Human:",
|
149 |
-
"[Your Name]",
|
150 |
-
"[Student]",
|
151 |
-
"Let me know if you need any clarification",
|
152 |
-
"I hope this helps",
|
153 |
-
"Feel free to ask",
|
154 |
-
"Can you provide",
|
155 |
-
"Would you like",
|
156 |
-
"Do you want",
|
157 |
-
"Let me know",
|
158 |
-
"Please let me know",
|
159 |
-
"Is there anything else",
|
160 |
-
"Do you have any questions",
|
161 |
-
"Sure!",
|
162 |
-
"Here are some examples:"
|
163 |
-
]
|
164 |
-
|
165 |
-
# First remove conversation starters from the end
|
166 |
-
for closure in closures:
|
167 |
-
if response.lower().endswith(closure.lower()):
|
168 |
-
response = response[:-(len(closure))].strip()
|
169 |
-
|
170 |
-
# Then remove any remaining conversation patterns
|
171 |
-
conversation_patterns = [
|
172 |
-
r"\?\s*$", # Questions at the end
|
173 |
-
r"Sure!.*$", # Responses starting with Sure!
|
174 |
-
r"Here are.*examples:?\s*$", # Incomplete example lists
|
175 |
-
r"Can you.*\?\s*$", # Questions starting with Can you
|
176 |
-
r"Would you.*\?\s*$", # Questions starting with Would you
|
177 |
-
r"Do you.*\?\s*$", # Questions starting with Do you
|
178 |
-
r"Let me know.*$", # Let me know phrases
|
179 |
-
r"I hope.*$", # I hope phrases
|
180 |
-
r"Feel free.*$" # Feel free phrases
|
181 |
-
]
|
182 |
-
|
183 |
-
import re
|
184 |
-
for pattern in conversation_patterns:
|
185 |
-
response = re.sub(pattern, "", response).strip()
|
186 |
-
|
187 |
-
print(f"After conversation removal: {response}") # Debug logging
|
188 |
|
189 |
# Ensure code examples are properly formatted
|
190 |
if "```python" not in response and "def " in response:
|
191 |
response = "```python\n" + response + "\n```"
|
192 |
|
193 |
-
#
|
194 |
-
if
|
195 |
-
|
196 |
-
response.strip().endswith("?") or
|
197 |
-
"can you" in response.lower() or
|
198 |
-
"let me know" in response.lower()):
|
199 |
-
print("Response validation failed - using fallback") # Debug logging
|
200 |
-
|
201 |
-
if "machine learning" in prompt.lower():
|
202 |
-
fallback_response = """Machine learning is a branch of artificial intelligence that enables computers to learn and improve from experience without being explicitly programmed. Think of it like teaching a child:
|
203 |
-
|
204 |
-
1. Simple Explanation:
|
205 |
-
- Instead of giving strict rules, we show the computer many examples
|
206 |
-
- The computer finds patterns in these examples
|
207 |
-
- It uses these patterns to make decisions about new situations
|
208 |
-
|
209 |
-
2. Real-World Applications:
|
210 |
-
- Email Spam Detection: Learning to identify unwanted emails based on past examples
|
211 |
-
- Netflix Recommendations: Suggesting movies based on what you've watched
|
212 |
-
- Face Recognition: Unlocking your phone by learning your facial features
|
213 |
-
- Virtual Assistants: Siri and Alexa understanding and responding to voice commands
|
214 |
-
- Medical Diagnosis: Helping doctors identify diseases in medical images
|
215 |
-
- Fraud Detection: Banks identifying suspicious transactions
|
216 |
-
|
217 |
-
3. Key Benefits:
|
218 |
-
- Automation of complex tasks
|
219 |
-
- More accurate predictions over time
|
220 |
-
- Ability to handle large amounts of data
|
221 |
-
- Continuous improvement through learning
|
222 |
-
|
223 |
-
Machine learning is transforming industries by automating tasks that once required human intelligence, making processes more efficient and enabling new possibilities in technology."""
|
224 |
-
elif "function" in prompt.lower():
|
225 |
fallback_response = """```python
|
226 |
def add_numbers(a, b):
|
227 |
-
'''
|
228 |
-
Add two numbers and return the result
|
229 |
-
Args:
|
230 |
-
a: first number
|
231 |
-
b: second number
|
232 |
-
Returns:
|
233 |
-
sum of a and b
|
234 |
-
'''
|
235 |
return a + b
|
236 |
-
|
237 |
-
# Example usage
|
238 |
-
num1 = 5
|
239 |
-
num2 = 3
|
240 |
-
result = add_numbers(num1, num2)
|
241 |
-
print(f"The sum of {num1} and {num2} is: {result}") # Output: The sum of 5 and 3 is: 8
|
242 |
```"""
|
243 |
else:
|
244 |
-
fallback_response = "I apologize, but I couldn't generate a
|
245 |
|
246 |
response = fallback_response
|
247 |
|
|
|
|
|
|
|
248 |
generation_time = time.time() - start_time
|
249 |
return response, generation_time
|
250 |
except Exception as e:
|
@@ -254,7 +160,7 @@ print(f"The sum of {num1} and {num2} is: {result}") # Output: The sum of 5 and
|
|
254 |
# Initialize model wrapper
|
255 |
model_wrapper = ModelWrapper()
|
256 |
|
257 |
-
def generate_text(prompt, max_length=
|
258 |
"""Gradio interface function"""
|
259 |
try:
|
260 |
if not prompt.strip():
|
@@ -269,91 +175,74 @@ def generate_text(prompt, max_length=384, temperature=0.7, top_p=0.9): # Reduce
|
|
269 |
return f"Generated in {gen_time:.2f} seconds:\n\n{response}"
|
270 |
except Exception as e:
|
271 |
print(f"Error in generate_text: {str(e)}")
|
272 |
-
return f"Error generating response: {str(e)}\nPlease try again with a
|
273 |
|
274 |
-
# Create
|
275 |
demo = gr.Interface(
|
276 |
fn=generate_text,
|
277 |
inputs=[
|
278 |
gr.Textbox(
|
279 |
label="Enter your prompt",
|
280 |
placeholder="Type your prompt here...",
|
281 |
-
lines=
|
282 |
),
|
283 |
gr.Slider(
|
284 |
minimum=64,
|
285 |
-
maximum=
|
286 |
-
value=
|
287 |
-
step=
|
288 |
label="Maximum Length",
|
289 |
-
info="
|
290 |
),
|
291 |
gr.Slider(
|
292 |
minimum=0.1,
|
293 |
-
maximum=
|
294 |
-
value=0.
|
295 |
step=0.1,
|
296 |
label="Temperature",
|
297 |
-
info="
|
298 |
),
|
299 |
gr.Slider(
|
300 |
-
minimum=0.
|
301 |
-
maximum=
|
302 |
-
value=0.
|
303 |
step=0.1,
|
304 |
label="Top P",
|
305 |
-
info="Controls diversity
|
306 |
),
|
307 |
],
|
308 |
-
outputs=gr.Textbox(label="Generated Response", lines=
|
309 |
-
title="Phi-2 QLoRA
|
310 |
-
description="""This is a
|
311 |
-
The model has been trained to provide helpful responses for various tasks including coding, writing, and general assistance.
|
312 |
-
|
313 |
-
Example tasks:
|
314 |
-
- Writing Python functions and explaining code
|
315 |
-
- Explaining technical concepts in simple terms
|
316 |
-
- Drafting professional emails and documents
|
317 |
|
318 |
Tips:
|
319 |
-
-
|
320 |
-
-
|
321 |
-
-
|
322 |
""",
|
323 |
examples=[
|
324 |
[
|
325 |
-
"Write a Python function to calculate
|
326 |
-
|
327 |
-
0.
|
328 |
-
0.
|
329 |
-
],
|
330 |
-
[
|
331 |
-
"Explain what machine learning is in simple terms and provide some real-world applications",
|
332 |
-
384,
|
333 |
-
0.5,
|
334 |
-
0.85
|
335 |
-
],
|
336 |
-
[
|
337 |
-
"Write a professional email to schedule a team meeting for next week to discuss project progress",
|
338 |
-
384,
|
339 |
-
0.5,
|
340 |
-
0.85
|
341 |
],
|
342 |
[
|
343 |
-
"
|
344 |
-
|
345 |
-
0.
|
346 |
-
0.
|
347 |
],
|
348 |
[
|
349 |
-
"
|
350 |
-
|
351 |
-
0.
|
352 |
-
0.
|
353 |
]
|
354 |
],
|
355 |
cache_examples=False
|
356 |
)
|
357 |
|
358 |
if __name__ == "__main__":
|
|
|
359 |
demo.launch()
|
|
|
5 |
import time
|
6 |
import gc
|
7 |
import os
|
8 |
+
import psutil
|
9 |
|
10 |
# Configuration
|
11 |
BASE_MODEL = "microsoft/phi-2"
|
12 |
ADAPTER_MODEL = "pradeep6kumar2024/phi2-qlora-assistant"
|
13 |
|
14 |
+
# Memory monitoring
|
15 |
+
def get_memory_usage():
|
16 |
+
process = psutil.Process(os.getpid())
|
17 |
+
return process.memory_info().rss / (1024 * 1024) # MB
|
18 |
+
|
19 |
class ModelWrapper:
|
20 |
def __init__(self):
|
21 |
self.model = None
|
|
|
32 |
# Clear memory
|
33 |
gc.collect()
|
34 |
|
35 |
+
print(f"Memory before loading: {get_memory_usage():.2f} MB")
|
36 |
+
|
37 |
print("Loading tokenizer...")
|
38 |
self.tokenizer = AutoTokenizer.from_pretrained(
|
39 |
BASE_MODEL,
|
|
|
42 |
)
|
43 |
self.tokenizer.pad_token = self.tokenizer.eos_token
|
44 |
|
45 |
+
print(f"Memory after tokenizer: {get_memory_usage():.2f} MB")
|
46 |
+
|
47 |
print("Loading base model...")
|
48 |
base_model = AutoModelForCausalLM.from_pretrained(
|
49 |
BASE_MODEL,
|
50 |
+
torch_dtype=torch.float32,
|
51 |
device_map="cpu",
|
52 |
trust_remote_code=True,
|
53 |
use_flash_attention_2=False,
|
54 |
+
low_cpu_mem_usage=True,
|
55 |
+
offload_folder="offload"
|
56 |
)
|
57 |
|
58 |
+
print(f"Memory after base model: {get_memory_usage():.2f} MB")
|
59 |
+
|
60 |
print("Loading LoRA adapter...")
|
61 |
self.model = PeftModel.from_pretrained(
|
62 |
base_model,
|
63 |
ADAPTER_MODEL,
|
64 |
+
torch_dtype=torch.float32,
|
65 |
device_map="cpu"
|
66 |
)
|
67 |
|
|
|
69 |
del base_model
|
70 |
gc.collect()
|
71 |
|
72 |
+
print(f"Memory after adapter: {get_memory_usage():.2f} MB")
|
73 |
+
|
74 |
self.model.eval()
|
75 |
print("Model loading complete!")
|
76 |
self.loaded = True
|
|
|
78 |
print(f"Error during model loading: {str(e)}")
|
79 |
raise
|
80 |
|
81 |
+
def generate_response(self, prompt, max_length=256, temperature=0.7, top_p=0.9):
|
82 |
if not self.loaded:
|
83 |
self.load_model()
|
84 |
|
85 |
try:
|
86 |
+
# Use shorter prompts to save memory
|
87 |
if "function" in prompt.lower() and "python" in prompt.lower():
|
88 |
+
enhanced_prompt = f"""Write Python function: {prompt}"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
89 |
elif any(word in prompt.lower() for word in ["explain", "what is", "how does", "describe"]):
|
90 |
+
enhanced_prompt = f"""Explain briefly: {prompt}"""
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
91 |
else:
|
92 |
+
enhanced_prompt = prompt
|
|
|
|
|
|
|
|
|
93 |
|
94 |
+
print(f"Enhanced prompt: {enhanced_prompt}")
|
95 |
|
96 |
+
# Tokenize input with shorter max length
|
97 |
inputs = self.tokenizer(
|
98 |
enhanced_prompt,
|
99 |
return_tensors="pt",
|
100 |
truncation=True,
|
101 |
+
max_length=256, # Reduced for memory
|
102 |
padding=True
|
103 |
+
).to("cpu")
|
104 |
|
105 |
+
# Generate with minimal parameters
|
106 |
start_time = time.time()
|
107 |
with torch.no_grad():
|
108 |
outputs = self.model.generate(
|
109 |
**inputs,
|
110 |
+
max_length=min(max_length, 256), # Strict limit
|
111 |
+
min_length=10, # Reduced minimum
|
112 |
temperature=min(0.5, temperature),
|
113 |
top_p=min(0.85, top_p),
|
114 |
do_sample=True,
|
115 |
pad_token_id=self.tokenizer.pad_token_id,
|
116 |
eos_token_id=self.tokenizer.eos_token_id,
|
117 |
+
repetition_penalty=1.2,
|
118 |
+
no_repeat_ngram_size=3,
|
119 |
num_return_sequences=1,
|
120 |
early_stopping=True,
|
121 |
+
num_beams=1, # Greedy decoding to save memory
|
122 |
length_penalty=0.6
|
123 |
)
|
124 |
|
125 |
# Decode response
|
126 |
response = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
|
|
127 |
|
128 |
# Clean up the response
|
129 |
if response.startswith(enhanced_prompt):
|
130 |
response = response[len(enhanced_prompt):].strip()
|
131 |
|
132 |
+
# Basic cleanup only
|
133 |
+
response = response.replace("Human:", "").replace("Assistant:", "")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
134 |
|
135 |
# Ensure code examples are properly formatted
|
136 |
if "```python" not in response and "def " in response:
|
137 |
response = "```python\n" + response + "\n```"
|
138 |
|
139 |
+
# Simple validation
|
140 |
+
if len(response.strip()) < 10:
|
141 |
+
if "function" in prompt.lower():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
142 |
fallback_response = """```python
|
143 |
def add_numbers(a, b):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
144 |
return a + b
|
|
|
|
|
|
|
|
|
|
|
|
|
145 |
```"""
|
146 |
else:
|
147 |
+
fallback_response = "I apologize, but I couldn't generate a response. Please try with a simpler prompt."
|
148 |
|
149 |
response = fallback_response
|
150 |
|
151 |
+
# Clear memory after generation
|
152 |
+
gc.collect()
|
153 |
+
|
154 |
generation_time = time.time() - start_time
|
155 |
return response, generation_time
|
156 |
except Exception as e:
|
|
|
160 |
# Initialize model wrapper
|
161 |
model_wrapper = ModelWrapper()
|
162 |
|
163 |
+
def generate_text(prompt, max_length=256, temperature=0.5, top_p=0.85):
|
164 |
"""Gradio interface function"""
|
165 |
try:
|
166 |
if not prompt.strip():
|
|
|
175 |
return f"Generated in {gen_time:.2f} seconds:\n\n{response}"
|
176 |
except Exception as e:
|
177 |
print(f"Error in generate_text: {str(e)}")
|
178 |
+
return f"Error generating response: {str(e)}\nPlease try again with a shorter prompt."
|
179 |
|
180 |
+
# Create a very lightweight Gradio interface
|
181 |
demo = gr.Interface(
|
182 |
fn=generate_text,
|
183 |
inputs=[
|
184 |
gr.Textbox(
|
185 |
label="Enter your prompt",
|
186 |
placeholder="Type your prompt here...",
|
187 |
+
lines=3
|
188 |
),
|
189 |
gr.Slider(
|
190 |
minimum=64,
|
191 |
+
maximum=256,
|
192 |
+
value=192,
|
193 |
+
step=32,
|
194 |
label="Maximum Length",
|
195 |
+
info="Keep this low for CPU"
|
196 |
),
|
197 |
gr.Slider(
|
198 |
minimum=0.1,
|
199 |
+
maximum=0.7,
|
200 |
+
value=0.4,
|
201 |
step=0.1,
|
202 |
label="Temperature",
|
203 |
+
info="Lower is better for CPU"
|
204 |
),
|
205 |
gr.Slider(
|
206 |
+
minimum=0.5,
|
207 |
+
maximum=0.9,
|
208 |
+
value=0.8,
|
209 |
step=0.1,
|
210 |
label="Top P",
|
211 |
+
info="Controls diversity"
|
212 |
),
|
213 |
],
|
214 |
+
outputs=gr.Textbox(label="Generated Response", lines=6),
|
215 |
+
title="Phi-2 QLoRA Assistant (CPU-Optimized)",
|
216 |
+
description="""This is a lightweight CPU version of the fine-tuned Phi-2 model.
|
|
|
|
|
|
|
|
|
|
|
|
|
217 |
|
218 |
Tips:
|
219 |
+
- Keep prompts short and specific
|
220 |
+
- Use lower maximum length (128-192) for faster responses
|
221 |
+
- Use lower temperature (0.3-0.5) for more reliable responses
|
222 |
""",
|
223 |
examples=[
|
224 |
[
|
225 |
+
"Write a Python function to calculate factorial",
|
226 |
+
192,
|
227 |
+
0.4,
|
228 |
+
0.8
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
229 |
],
|
230 |
[
|
231 |
+
"Explain machine learning simply",
|
232 |
+
192,
|
233 |
+
0.4,
|
234 |
+
0.8
|
235 |
],
|
236 |
[
|
237 |
+
"Write a short email to schedule a meeting",
|
238 |
+
192,
|
239 |
+
0.4,
|
240 |
+
0.8
|
241 |
]
|
242 |
],
|
243 |
cache_examples=False
|
244 |
)
|
245 |
|
246 |
if __name__ == "__main__":
|
247 |
+
demo.queue(concurrency_count=1) # Limit concurrency
|
248 |
demo.launch()
|