Spaces:
Runtime error
Runtime error
Arjun Moorthy
commited on
Commit
Β·
2720b05
1
Parent(s):
da47961
Optimize for hardware constraints - make RAG optional and lightweight
Browse files- Oncolife/app.py +59 -52
- requirements.txt +1 -11
Oncolife/app.py
CHANGED
@@ -4,7 +4,7 @@ OncoLife Symptom & Triage Assistant
|
|
4 |
A medical chatbot that performs both symptom assessment and clinical triage for chemotherapy patients.
|
5 |
Updated: Using BioMistral-7B base model for medical conversations.
|
6 |
REBUILD: Simplified to use only base model, no adapters.
|
7 |
-
RAG: Added document retrieval capabilities for PDFs and other reference materials.
|
8 |
"""
|
9 |
|
10 |
import gradio as gr
|
@@ -15,14 +15,19 @@ from transformers import AutoTokenizer, MistralForCausalLM
|
|
15 |
import torch
|
16 |
from spaces import GPU
|
17 |
|
18 |
-
# RAG imports
|
19 |
-
|
20 |
-
|
21 |
-
import
|
22 |
-
import
|
23 |
-
|
24 |
-
from langchain.
|
25 |
-
|
|
|
|
|
|
|
|
|
|
|
26 |
|
27 |
# Force GPU detection for HF Spaces
|
28 |
@GPU
|
@@ -51,8 +56,18 @@ class OncoLifeAssistant:
|
|
51 |
# Load the OncoLife instructions
|
52 |
self._load_instructions()
|
53 |
|
54 |
-
# Initialize RAG system
|
55 |
-
self.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
56 |
|
57 |
def _load_instructions(self):
|
58 |
"""Load the OncoLife instructions from the text file"""
|
@@ -70,15 +85,15 @@ class OncoLifeAssistant:
|
|
70 |
self.instructions = ""
|
71 |
|
72 |
def _initialize_rag(self):
|
73 |
-
"""Initialize the RAG system with document embeddings"""
|
74 |
try:
|
75 |
-
print("π Initializing RAG system...")
|
76 |
|
77 |
-
#
|
78 |
self.embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
79 |
print("β
Loaded embedding model")
|
80 |
|
81 |
-
# Initialize ChromaDB
|
82 |
self.chroma_client = chromadb.Client()
|
83 |
self.collection = self.chroma_client.create_collection(
|
84 |
name="oncolife_documents",
|
@@ -86,19 +101,20 @@ class OncoLifeAssistant:
|
|
86 |
)
|
87 |
print("β
Initialized ChromaDB collection")
|
88 |
|
89 |
-
# Load and process documents
|
90 |
-
self.
|
91 |
|
92 |
except Exception as e:
|
93 |
print(f"β Error initializing RAG: {e}")
|
94 |
self.embedding_model = None
|
95 |
self.collection = None
|
|
|
96 |
|
97 |
-
def
|
98 |
-
"""Load
|
99 |
try:
|
100 |
docs_path = Path(__file__).parent / "guideline-docs"
|
101 |
-
print(f"π Loading documents from: {docs_path}")
|
102 |
|
103 |
if not docs_path.exists():
|
104 |
print("β οΈ guideline-docs directory not found")
|
@@ -106,27 +122,14 @@ class OncoLifeAssistant:
|
|
106 |
|
107 |
# Text splitter for chunking documents
|
108 |
text_splitter = RecursiveCharacterTextSplitter(
|
109 |
-
chunk_size=
|
110 |
-
chunk_overlap=
|
111 |
separators=["\n\n", "\n", ". ", " ", ""]
|
112 |
)
|
113 |
|
114 |
documents_loaded = 0
|
115 |
|
116 |
-
#
|
117 |
-
for pdf_file in docs_path.glob("*.pdf"):
|
118 |
-
try:
|
119 |
-
print(f"π Processing PDF: {pdf_file.name}")
|
120 |
-
text = self._extract_pdf_text(pdf_file)
|
121 |
-
if text:
|
122 |
-
chunks = text_splitter.split_text(text)
|
123 |
-
self._add_chunks_to_db(chunks, pdf_file.name)
|
124 |
-
documents_loaded += 1
|
125 |
-
print(f"β
Added {len(chunks)} chunks from {pdf_file.name}")
|
126 |
-
except Exception as e:
|
127 |
-
print(f"β Error processing {pdf_file.name}: {e}")
|
128 |
-
|
129 |
-
# Process JSON files
|
130 |
for json_file in docs_path.glob("*.json"):
|
131 |
try:
|
132 |
print(f"π Processing JSON: {json_file.name}")
|
@@ -141,7 +144,7 @@ class OncoLifeAssistant:
|
|
141 |
except Exception as e:
|
142 |
print(f"β Error processing {json_file.name}: {e}")
|
143 |
|
144 |
-
# Process text files
|
145 |
for txt_file in docs_path.glob("*.txt"):
|
146 |
try:
|
147 |
print(f"π Processing TXT: {txt_file.name}")
|
@@ -222,10 +225,10 @@ class OncoLifeAssistant:
|
|
222 |
except Exception as e:
|
223 |
print(f"β Error adding chunks to database: {e}")
|
224 |
|
225 |
-
def _retrieve_relevant_documents(self, query, top_k=
|
226 |
"""Retrieve relevant document chunks for a query"""
|
227 |
try:
|
228 |
-
if not self.collection or not self.embedding_model:
|
229 |
return []
|
230 |
|
231 |
# Generate query embedding
|
@@ -254,7 +257,7 @@ class OncoLifeAssistant:
|
|
254 |
return []
|
255 |
|
256 |
def _load_model(self, model_id, gpu_available):
|
257 |
-
"""Load the BioMistral base model"""
|
258 |
try:
|
259 |
print("π Loading BioMistral base model...")
|
260 |
|
@@ -275,14 +278,16 @@ class OncoLifeAssistant:
|
|
275 |
trust_remote_code=True
|
276 |
)
|
277 |
|
278 |
-
# Load the model
|
279 |
print(f"π¦ Loading model: {model_id}")
|
280 |
self.model = MistralForCausalLM.from_pretrained(
|
281 |
model_id,
|
282 |
trust_remote_code=True,
|
283 |
device_map="auto",
|
284 |
torch_dtype=dtype,
|
285 |
-
low_cpu_mem_usage=True
|
|
|
|
|
286 |
)
|
287 |
|
288 |
# Add pad token if not present
|
@@ -297,7 +302,7 @@ class OncoLifeAssistant:
|
|
297 |
self.tokenizer = None
|
298 |
|
299 |
def generate_oncolife_response(self, user_input, conversation_history):
|
300 |
-
"""Generate response using OncoLife instructions and RAG"""
|
301 |
try:
|
302 |
if self.model is None or self.tokenizer is None:
|
303 |
return """β **Model Loading Error**
|
@@ -311,15 +316,17 @@ Please check the Space logs for details."""
|
|
311 |
|
312 |
print(f"π Generating OncoLife response for: {user_input}")
|
313 |
|
314 |
-
# Retrieve relevant documents using RAG
|
315 |
-
relevant_docs = self._retrieve_relevant_documents(user_input, top_k=3)
|
316 |
-
|
317 |
-
# Format retrieved documents
|
318 |
context_text = ""
|
319 |
-
if
|
320 |
-
|
321 |
-
|
322 |
-
|
|
|
|
|
|
|
|
|
|
|
323 |
|
324 |
# Create prompt using the loaded instructions and retrieved context
|
325 |
system_prompt = f"""You are the OncoLife Symptom & Triage Assistant. Follow these instructions exactly:
|
@@ -426,7 +433,7 @@ Please try a simpler question or check the logs for more details."""
|
|
426 |
"assistant": assistant_msg
|
427 |
})
|
428 |
|
429 |
-
# Generate response using OncoLife instructions and RAG
|
430 |
response = self.generate_oncolife_response(message, conversation_history)
|
431 |
|
432 |
return response
|
|
|
4 |
A medical chatbot that performs both symptom assessment and clinical triage for chemotherapy patients.
|
5 |
Updated: Using BioMistral-7B base model for medical conversations.
|
6 |
REBUILD: Simplified to use only base model, no adapters.
|
7 |
+
RAG: Added document retrieval capabilities for PDFs and other reference materials (optional).
|
8 |
"""
|
9 |
|
10 |
import gradio as gr
|
|
|
15 |
import torch
|
16 |
from spaces import GPU
|
17 |
|
18 |
+
# RAG imports (optional)
|
19 |
+
try:
|
20 |
+
import chromadb
|
21 |
+
from sentence_transformers import SentenceTransformer
|
22 |
+
import PyPDF2
|
23 |
+
import pdfplumber
|
24 |
+
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
25 |
+
from langchain.embeddings import HuggingFaceEmbeddings
|
26 |
+
import fitz # PyMuPDF for better PDF handling
|
27 |
+
RAG_AVAILABLE = True
|
28 |
+
except ImportError:
|
29 |
+
print("β οΈ RAG libraries not available, running in instruction-only mode")
|
30 |
+
RAG_AVAILABLE = False
|
31 |
|
32 |
# Force GPU detection for HF Spaces
|
33 |
@GPU
|
|
|
56 |
# Load the OncoLife instructions
|
57 |
self._load_instructions()
|
58 |
|
59 |
+
# Initialize RAG system (optional)
|
60 |
+
self.rag_enabled = False
|
61 |
+
if RAG_AVAILABLE:
|
62 |
+
try:
|
63 |
+
self._initialize_rag()
|
64 |
+
self.rag_enabled = True
|
65 |
+
print("β
RAG system initialized successfully")
|
66 |
+
except Exception as e:
|
67 |
+
print(f"β οΈ RAG initialization failed: {e}")
|
68 |
+
print("π Continuing with instruction-only mode")
|
69 |
+
else:
|
70 |
+
print("π Running in instruction-only mode (no RAG)")
|
71 |
|
72 |
def _load_instructions(self):
|
73 |
"""Load the OncoLife instructions from the text file"""
|
|
|
85 |
self.instructions = ""
|
86 |
|
87 |
def _initialize_rag(self):
|
88 |
+
"""Initialize the RAG system with document embeddings (lightweight version)"""
|
89 |
try:
|
90 |
+
print("π Initializing lightweight RAG system...")
|
91 |
|
92 |
+
# Use a smaller embedding model
|
93 |
self.embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
|
94 |
print("β
Loaded embedding model")
|
95 |
|
96 |
+
# Initialize ChromaDB with persistence disabled for memory efficiency
|
97 |
self.chroma_client = chromadb.Client()
|
98 |
self.collection = self.chroma_client.create_collection(
|
99 |
name="oncolife_documents",
|
|
|
101 |
)
|
102 |
print("β
Initialized ChromaDB collection")
|
103 |
|
104 |
+
# Load and process documents (limited to essential files)
|
105 |
+
self._load_documents_lightweight()
|
106 |
|
107 |
except Exception as e:
|
108 |
print(f"β Error initializing RAG: {e}")
|
109 |
self.embedding_model = None
|
110 |
self.collection = None
|
111 |
+
raise e
|
112 |
|
113 |
+
def _load_documents_lightweight(self):
|
114 |
+
"""Load only essential documents to save memory"""
|
115 |
try:
|
116 |
docs_path = Path(__file__).parent / "guideline-docs"
|
117 |
+
print(f"π Loading essential documents from: {docs_path}")
|
118 |
|
119 |
if not docs_path.exists():
|
120 |
print("β οΈ guideline-docs directory not found")
|
|
|
122 |
|
123 |
# Text splitter for chunking documents
|
124 |
text_splitter = RecursiveCharacterTextSplitter(
|
125 |
+
chunk_size=500, # Smaller chunks to save memory
|
126 |
+
chunk_overlap=100,
|
127 |
separators=["\n\n", "\n", ". ", " ", ""]
|
128 |
)
|
129 |
|
130 |
documents_loaded = 0
|
131 |
|
132 |
+
# Only process JSON files (lightweight)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
133 |
for json_file in docs_path.glob("*.json"):
|
134 |
try:
|
135 |
print(f"π Processing JSON: {json_file.name}")
|
|
|
144 |
except Exception as e:
|
145 |
print(f"β Error processing {json_file.name}: {e}")
|
146 |
|
147 |
+
# Process text files (lightweight)
|
148 |
for txt_file in docs_path.glob("*.txt"):
|
149 |
try:
|
150 |
print(f"π Processing TXT: {txt_file.name}")
|
|
|
225 |
except Exception as e:
|
226 |
print(f"β Error adding chunks to database: {e}")
|
227 |
|
228 |
+
def _retrieve_relevant_documents(self, query, top_k=3):
|
229 |
"""Retrieve relevant document chunks for a query"""
|
230 |
try:
|
231 |
+
if not self.collection or not self.embedding_model or not self.rag_enabled:
|
232 |
return []
|
233 |
|
234 |
# Generate query embedding
|
|
|
257 |
return []
|
258 |
|
259 |
def _load_model(self, model_id, gpu_available):
|
260 |
+
"""Load the BioMistral base model with memory optimization"""
|
261 |
try:
|
262 |
print("π Loading BioMistral base model...")
|
263 |
|
|
|
278 |
trust_remote_code=True
|
279 |
)
|
280 |
|
281 |
+
# Load the model with memory optimization
|
282 |
print(f"π¦ Loading model: {model_id}")
|
283 |
self.model = MistralForCausalLM.from_pretrained(
|
284 |
model_id,
|
285 |
trust_remote_code=True,
|
286 |
device_map="auto",
|
287 |
torch_dtype=dtype,
|
288 |
+
low_cpu_mem_usage=True,
|
289 |
+
# Add memory optimization
|
290 |
+
max_memory={0: "8GB", "cpu": "16GB"} if gpu_available else {"cpu": "8GB"}
|
291 |
)
|
292 |
|
293 |
# Add pad token if not present
|
|
|
302 |
self.tokenizer = None
|
303 |
|
304 |
def generate_oncolife_response(self, user_input, conversation_history):
|
305 |
+
"""Generate response using OncoLife instructions and optional RAG"""
|
306 |
try:
|
307 |
if self.model is None or self.tokenizer is None:
|
308 |
return """β **Model Loading Error**
|
|
|
316 |
|
317 |
print(f"π Generating OncoLife response for: {user_input}")
|
318 |
|
319 |
+
# Retrieve relevant documents using RAG (if available)
|
|
|
|
|
|
|
320 |
context_text = ""
|
321 |
+
if self.rag_enabled:
|
322 |
+
try:
|
323 |
+
relevant_docs = self._retrieve_relevant_documents(user_input, top_k=2)
|
324 |
+
if relevant_docs:
|
325 |
+
context_text = "\n\n**Relevant Reference Information:**\n"
|
326 |
+
for i, doc in enumerate(relevant_docs):
|
327 |
+
context_text += f"\n--- Source: {doc['source']} ---\n{doc['content'][:300]}...\n"
|
328 |
+
except Exception as e:
|
329 |
+
print(f"β οΈ RAG retrieval failed: {e}")
|
330 |
|
331 |
# Create prompt using the loaded instructions and retrieved context
|
332 |
system_prompt = f"""You are the OncoLife Symptom & Triage Assistant. Follow these instructions exactly:
|
|
|
433 |
"assistant": assistant_msg
|
434 |
})
|
435 |
|
436 |
+
# Generate response using OncoLife instructions and optional RAG
|
437 |
response = self.generate_oncolife_response(message, conversation_history)
|
438 |
|
439 |
return response
|
requirements.txt
CHANGED
@@ -1,5 +1,3 @@
|
|
1 |
-
# Medical Chatbot HF Space Requirements
|
2 |
-
|
3 |
# Web framework
|
4 |
gradio==4.44.0
|
5 |
|
@@ -9,18 +7,10 @@ transformers==4.36.2
|
|
9 |
accelerate==0.25.0
|
10 |
|
11 |
# HF Spaces GPU support
|
12 |
-
spaces>=0.1.0
|
13 |
-
|
14 |
-
# Basic utilities
|
15 |
-
numpy>=1.21.0,<2.0.0
|
16 |
-
requests>=2.28.0
|
17 |
-
|
18 |
-
# Additional dependencies for better device handling
|
19 |
safetensors==0.4.1
|
20 |
tokenizers>=0.15.0
|
21 |
|
22 |
-
# RAG implementation
|
23 |
-
bitsandbytes==0.41.3
|
24 |
sentence-transformers==2.2.2
|
25 |
chromadb==0.4.22
|
26 |
pypdf2==3.0.1
|
|
|
|
|
|
|
1 |
# Web framework
|
2 |
gradio==4.44.0
|
3 |
|
|
|
7 |
accelerate==0.25.0
|
8 |
|
9 |
# HF Spaces GPU support
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
10 |
safetensors==0.4.1
|
11 |
tokenizers>=0.15.0
|
12 |
|
13 |
+
# RAG implementation (optional - will fallback gracefully if not available)
|
|
|
14 |
sentence-transformers==2.2.2
|
15 |
chromadb==0.4.22
|
16 |
pypdf2==3.0.1
|