Update app.py
Browse files
app.py
CHANGED
@@ -1,6 +1,6 @@
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
-
from transformers import AutoTokenizer,
|
4 |
import logging
|
5 |
from typing import List, Dict
|
6 |
import gc
|
@@ -13,232 +13,113 @@ logging.basicConfig(
|
|
13 |
)
|
14 |
logger = logging.getLogger(__name__)
|
15 |
|
16 |
-
# Set
|
17 |
-
torch.
|
18 |
|
19 |
class HealthAssistant:
|
20 |
-
def __init__(self
|
21 |
-
|
22 |
-
self.model_name = "Qwen/Qwen2-VL-7B-Instruct"
|
23 |
-
else:
|
24 |
-
self.model_name = "Qwen/Qwen2-VL-7B-Instruct"
|
25 |
-
|
26 |
self.model = None
|
27 |
self.tokenizer = None
|
|
|
28 |
self.metrics = []
|
29 |
self.medications = []
|
30 |
self.initialize_model()
|
31 |
|
32 |
def initialize_model(self):
|
33 |
try:
|
34 |
-
logger.info(f"
|
35 |
|
|
|
36 |
self.tokenizer = AutoTokenizer.from_pretrained(
|
37 |
-
self.
|
38 |
trust_remote_code=True
|
39 |
)
|
40 |
logger.info("Tokenizer loaded")
|
41 |
|
|
|
42 |
self.model = AutoModelForCausalLM.from_pretrained(
|
43 |
-
self.
|
44 |
-
torch_dtype=
|
45 |
-
low_cpu_mem_usage=True,
|
46 |
trust_remote_code=True
|
47 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
-
if self.tokenizer.pad_token is None:
|
50 |
-
self.tokenizer.pad_token = self.tokenizer.eos_token
|
51 |
-
|
52 |
-
self.model = self.model.to("cpu")
|
53 |
-
logger.info("Model loaded successfully")
|
54 |
return True
|
55 |
|
56 |
except Exception as e:
|
57 |
logger.error(f"Error in model initialization: {str(e)}")
|
58 |
raise
|
59 |
|
60 |
-
def
|
61 |
-
"""
|
62 |
-
|
63 |
-
|
64 |
-
|
65 |
-
|
66 |
-
|
67 |
-
|
|
|
68 |
|
69 |
-
|
70 |
-
|
71 |
-
|
|
|
|
|
|
|
72 |
|
73 |
-
|
74 |
-
|
75 |
-
|
|
|
76 |
|
77 |
-
|
78 |
-
if any(keyword in message_lower for keyword in lifestyle_keywords):
|
79 |
-
return "lifestyle_advice"
|
80 |
-
|
81 |
-
return "general"
|
82 |
-
|
83 |
-
def _prepare_medical_prompt(self, message: str, query_type: str) -> str:
|
84 |
-
"""Prepare medical prompt based on query type"""
|
85 |
-
base_context = self._get_health_context()
|
86 |
-
|
87 |
-
prompts = {
|
88 |
-
"symptom_check": f"""You are a medical AI assistant. Based on the following health context and symptoms, provide a careful analysis.
|
89 |
-
|
90 |
-
Current Health Context:
|
91 |
-
{base_context}
|
92 |
-
|
93 |
-
Patient's Symptoms: {message}
|
94 |
-
|
95 |
-
Provide a structured response covering:
|
96 |
-
1. Key symptoms identified
|
97 |
-
2. Possible common causes
|
98 |
-
3. General recommendations
|
99 |
-
4. Warning signs to watch for
|
100 |
-
5. When to seek medical care
|
101 |
-
|
102 |
-
Remember to maintain a professional and careful tone.""",
|
103 |
-
|
104 |
-
"medication_info": f"""You are a medical AI assistant. Provide information about the medication inquiry while noting you cannot give prescription advice.
|
105 |
-
|
106 |
-
Current Health Context:
|
107 |
-
{base_context}
|
108 |
-
|
109 |
-
Medication Query: {message}
|
110 |
-
|
111 |
-
Provide general information about:
|
112 |
-
1. Basic medication category/purpose
|
113 |
-
2. General usage patterns
|
114 |
-
3. Common considerations
|
115 |
-
4. Important precautions
|
116 |
-
5. When to consult a healthcare provider
|
117 |
-
|
118 |
-
Remember to emphasize this is general information only.""",
|
119 |
-
|
120 |
-
"emergency_guidance": f"""You are a medical AI assistant. This appears to be an urgent situation.
|
121 |
-
|
122 |
-
Current Health Context:
|
123 |
-
{base_context}
|
124 |
-
|
125 |
-
Urgent Situation: {message}
|
126 |
-
|
127 |
-
Provide immediate guidance:
|
128 |
-
1. Severity assessment
|
129 |
-
2. Immediate actions needed
|
130 |
-
3. Emergency warning signs
|
131 |
-
4. Whether to call emergency services
|
132 |
-
5. Precautions while waiting
|
133 |
-
|
134 |
-
Always emphasize seeking immediate medical care for emergencies.""",
|
135 |
-
|
136 |
-
"general": f"""You are a medical AI assistant. Provide helpful health information based on the query.
|
137 |
-
|
138 |
-
Current Health Context:
|
139 |
-
{base_context}
|
140 |
-
|
141 |
-
Health Query: {message}
|
142 |
-
|
143 |
-
Provide a structured response covering:
|
144 |
-
1. Understanding of the question
|
145 |
-
2. Relevant health information
|
146 |
-
3. General guidance
|
147 |
-
4. Important considerations
|
148 |
-
5. Additional recommendations"""
|
149 |
-
}
|
150 |
-
|
151 |
-
return prompts.get(query_type, prompts["general"])
|
152 |
|
153 |
def generate_response(self, message: str, history: List = None) -> str:
|
154 |
try:
|
155 |
-
if not hasattr(self, 'model') or self.model is None:
|
156 |
-
return "System is initializing. Please try again in a moment."
|
157 |
-
|
158 |
-
# Detect query type
|
159 |
-
query_type = self._detect_query_type(message)
|
160 |
-
|
161 |
# Prepare prompt
|
162 |
-
prompt = self.
|
163 |
-
|
164 |
-
#
|
165 |
-
|
166 |
-
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
-
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
-
|
177 |
-
|
178 |
-
|
179 |
-
# Generate
|
180 |
-
with torch.no_grad():
|
181 |
-
outputs = self.model.generate(
|
182 |
-
inputs["input_ids"],
|
183 |
-
max_new_tokens=150,
|
184 |
-
num_beams=1,
|
185 |
-
temperature=0.7,
|
186 |
-
top_p=0.9,
|
187 |
-
pad_token_id=self.tokenizer.pad_token_id,
|
188 |
-
eos_token_id=self.tokenizer.eos_token_id
|
189 |
-
)
|
190 |
-
|
191 |
-
# Decode
|
192 |
-
response = self.tokenizer.decode(
|
193 |
-
outputs[0][inputs["input_ids"].shape[1]:],
|
194 |
-
skip_special_tokens=True
|
195 |
-
)
|
196 |
-
|
197 |
-
# Format response
|
198 |
-
response = self._format_response(response, query_type)
|
199 |
|
200 |
# Cleanup
|
201 |
-
del outputs, inputs
|
202 |
gc.collect()
|
|
|
|
|
203 |
|
204 |
return response.strip()
|
205 |
|
206 |
except Exception as e:
|
207 |
logger.error(f"Error generating response: {str(e)}")
|
208 |
-
return "I apologize, but I encountered an error. Please try
|
209 |
-
|
210 |
-
def _format_response(self, response: str, query_type: str) -> str:
|
211 |
-
"""Format and clean the response"""
|
212 |
-
# Remove repeated headers
|
213 |
-
lines = [line.strip() for line in response.split('\n') if line.strip()]
|
214 |
-
clean_lines = []
|
215 |
-
seen = set()
|
216 |
-
|
217 |
-
for line in lines:
|
218 |
-
if line not in seen:
|
219 |
-
seen.add(line)
|
220 |
-
clean_lines.append(line)
|
221 |
-
|
222 |
-
# Add appropriate prefix based on query type
|
223 |
-
prefixes = {
|
224 |
-
"emergency_guidance": "🚨 URGENT: ",
|
225 |
-
"symptom_check": "🔍 Analysis: ",
|
226 |
-
"medication_info": "💊 Medication Info: ",
|
227 |
-
"lifestyle_advice": "💡 Health Advice: ",
|
228 |
-
"general": "ℹ️ "
|
229 |
-
}
|
230 |
-
|
231 |
-
prefix = prefixes.get(query_type, "ℹ️ ")
|
232 |
-
formatted_response = prefix + "\n".join(clean_lines)
|
233 |
-
|
234 |
-
# Add disclaimer for certain types
|
235 |
-
if query_type in ["emergency_guidance", "medication_info"]:
|
236 |
-
formatted_response += "\n\n⚠️ Note: This is general information only. Always consult healthcare professionals."
|
237 |
-
|
238 |
-
return formatted_response
|
239 |
|
240 |
def _get_health_context(self) -> str:
|
241 |
-
"""Get user's health context"""
|
242 |
context_parts = []
|
243 |
|
244 |
if self.metrics:
|
@@ -289,7 +170,7 @@ class GradioInterface:
|
|
289 |
def __init__(self):
|
290 |
try:
|
291 |
logger.info("Initializing Health Assistant...")
|
292 |
-
self.assistant = HealthAssistant(
|
293 |
logger.info("Health Assistant initialized successfully")
|
294 |
except Exception as e:
|
295 |
logger.error(f"Failed to initialize Health Assistant: {e}")
|
|
|
1 |
import gradio as gr
|
2 |
import torch
|
3 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
|
4 |
import logging
|
5 |
from typing import List, Dict
|
6 |
import gc
|
|
|
13 |
)
|
14 |
logger = logging.getLogger(__name__)
|
15 |
|
16 |
+
# Set random seed for reproducibility
|
17 |
+
torch.random.manual_seed(0)
|
18 |
|
19 |
class HealthAssistant:
|
20 |
+
def __init__(self):
|
21 |
+
self.model_id = "microsoft/Phi-3-small-128k-instruct"
|
|
|
|
|
|
|
|
|
22 |
self.model = None
|
23 |
self.tokenizer = None
|
24 |
+
self.pipe = None
|
25 |
self.metrics = []
|
26 |
self.medications = []
|
27 |
self.initialize_model()
|
28 |
|
29 |
def initialize_model(self):
|
30 |
try:
|
31 |
+
logger.info(f"Loading model: {self.model_id}")
|
32 |
|
33 |
+
# Initialize tokenizer
|
34 |
self.tokenizer = AutoTokenizer.from_pretrained(
|
35 |
+
self.model_id,
|
36 |
trust_remote_code=True
|
37 |
)
|
38 |
logger.info("Tokenizer loaded")
|
39 |
|
40 |
+
# Initialize model
|
41 |
self.model = AutoModelForCausalLM.from_pretrained(
|
42 |
+
self.model_id,
|
43 |
+
torch_dtype="auto",
|
|
|
44 |
trust_remote_code=True
|
45 |
)
|
46 |
+
|
47 |
+
# Set device
|
48 |
+
self.device = "cuda" if torch.cuda.is_available() else "cpu"
|
49 |
+
self.model = self.model.to(self.device)
|
50 |
+
logger.info(f"Model loaded on {self.device}")
|
51 |
+
|
52 |
+
# Setup pipeline
|
53 |
+
self.pipe = pipeline(
|
54 |
+
"text-generation",
|
55 |
+
model=self.model,
|
56 |
+
tokenizer=self.tokenizer,
|
57 |
+
device=self.device
|
58 |
+
)
|
59 |
+
logger.info("Pipeline created successfully")
|
60 |
|
|
|
|
|
|
|
|
|
|
|
61 |
return True
|
62 |
|
63 |
except Exception as e:
|
64 |
logger.error(f"Error in model initialization: {str(e)}")
|
65 |
raise
|
66 |
|
67 |
+
def _prepare_prompt(self, message: str, history: List = None) -> str:
|
68 |
+
"""Prepare prompt with context and history"""
|
69 |
+
prompt_parts = [
|
70 |
+
"You are a medical AI assistant providing healthcare information and guidance.",
|
71 |
+
"Always be professional and include appropriate medical disclaimers.",
|
72 |
+
"\nCurrent Health Information:",
|
73 |
+
self._get_health_context(),
|
74 |
+
"\nConversation:"
|
75 |
+
]
|
76 |
|
77 |
+
if history:
|
78 |
+
for prev_msg, prev_response in history[-3:]:
|
79 |
+
prompt_parts.extend([
|
80 |
+
f"Human: {prev_msg}",
|
81 |
+
f"Assistant: {prev_response}"
|
82 |
+
])
|
83 |
|
84 |
+
prompt_parts.extend([
|
85 |
+
f"Human: {message}",
|
86 |
+
"Assistant:"
|
87 |
+
])
|
88 |
|
89 |
+
return "\n".join(prompt_parts)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
90 |
|
91 |
def generate_response(self, message: str, history: List = None) -> str:
|
92 |
try:
|
|
|
|
|
|
|
|
|
|
|
|
|
93 |
# Prepare prompt
|
94 |
+
prompt = self._prepare_prompt(message, history)
|
95 |
+
|
96 |
+
# Generation configuration
|
97 |
+
generation_args = {
|
98 |
+
"max_new_tokens": 500,
|
99 |
+
"return_full_text": False,
|
100 |
+
"temperature": 0.7,
|
101 |
+
"do_sample": True,
|
102 |
+
"top_k": 50,
|
103 |
+
"top_p": 0.9,
|
104 |
+
"repetition_penalty": 1.1
|
105 |
+
}
|
106 |
+
|
107 |
+
# Generate response
|
108 |
+
output = self.pipe(prompt, **generation_args)
|
109 |
+
response = output[0]['generated_text']
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
110 |
|
111 |
# Cleanup
|
|
|
112 |
gc.collect()
|
113 |
+
if torch.cuda.is_available():
|
114 |
+
torch.cuda.empty_cache()
|
115 |
|
116 |
return response.strip()
|
117 |
|
118 |
except Exception as e:
|
119 |
logger.error(f"Error generating response: {str(e)}")
|
120 |
+
return "I apologize, but I encountered an error. Please try again."
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
121 |
|
122 |
def _get_health_context(self) -> str:
|
|
|
123 |
context_parts = []
|
124 |
|
125 |
if self.metrics:
|
|
|
170 |
def __init__(self):
|
171 |
try:
|
172 |
logger.info("Initializing Health Assistant...")
|
173 |
+
self.assistant = HealthAssistant()
|
174 |
logger.info("Health Assistant initialized successfully")
|
175 |
except Exception as e:
|
176 |
logger.error(f"Failed to initialize Health Assistant: {e}")
|