Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -8,22 +8,17 @@ from enum import Enum
|
|
8 |
import gradio as gr
|
9 |
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
10 |
from langchain_community.vectorstores import Chroma
|
11 |
-
from langchain.prompts import PromptTemplate
|
12 |
from langchain.schema import BaseRetriever
|
13 |
from langchain.embeddings.base import Embeddings
|
14 |
from langchain.llms.base import BaseLanguageModel
|
15 |
import PyPDF2
|
|
|
16 |
# Install required packages
|
17 |
|
18 |
|
19 |
# Initialize models
|
20 |
import torch
|
21 |
-
from transformers import AutoModelForCausalLM, AutoTokenizer, BitsAndBytesConfig
|
22 |
from langchain_community.embeddings import HuggingFaceBgeEmbeddings
|
23 |
-
from langchain_community.llms import HuggingFacePipeline
|
24 |
-
from transformers import pipeline
|
25 |
-
from sentence_transformers import SentenceTransformer
|
26 |
-
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
27 |
embed_model = HuggingFaceBgeEmbeddings(
|
28 |
model_name="all-MiniLM-L6-v2",#"dunzhang/stella_en_1.5B_v5",
|
29 |
model_kwargs={'device': 'cpu'},
|
@@ -31,34 +26,10 @@ embed_model = HuggingFaceBgeEmbeddings(
|
|
31 |
)
|
32 |
|
33 |
model_name = "meta-llama/Llama-3.2-3B-Instruct"#"google/gemma-2-2b-it"#"prithivMLmods/Llama-3.2-3B-GGUF"
|
34 |
-
from huggingface_hub import InferenceClient
|
35 |
-
|
36 |
-
client = InferenceClient(model_name)
|
37 |
|
38 |
|
|
|
39 |
|
40 |
-
# tokenizer = AutoTokenizer.from_pretrained(model_name)
|
41 |
-
# model = AutoModelForCausalLM.from_pretrained(
|
42 |
-
# model_name,
|
43 |
-
# trust_remote_code=True,
|
44 |
-
# use_auth_token=True
|
45 |
-
# )
|
46 |
-
|
47 |
-
# pipe = pipeline(
|
48 |
-
# "text-generation",
|
49 |
-
# model=model,
|
50 |
-
# tokenizer=tokenizer,
|
51 |
-
# max_new_tokens=2048*2,
|
52 |
-
# temperature=0.3,
|
53 |
-
# top_p=0.95,
|
54 |
-
# generation_config=model.generation_config
|
55 |
-
# # repetition_penalty=1.15
|
56 |
-
# )
|
57 |
-
# llm = HuggingFacePipeline(pipeline=pipe)
|
58 |
-
# model.generation_config.pad_token_id = model.generation_config.eos_token_id
|
59 |
-
|
60 |
-
|
61 |
-
# embed_model = embedding_model
|
62 |
|
63 |
# Set up logging
|
64 |
logging.basicConfig(level=logging.INFO)
|
@@ -71,24 +42,14 @@ class DocumentFormat(Enum):
|
|
71 |
@dataclass
|
72 |
class RAGConfig:
|
73 |
"""Configuration for RAG system parameters"""
|
74 |
-
chunk_size: int =
|
75 |
-
chunk_overlap: int =
|
76 |
retriever_k: int = 3
|
77 |
persist_directory: str = "./chroma_db"
|
78 |
|
79 |
class AdvancedRAGSystem:
|
80 |
"""Advanced RAG System with improved error handling and type safety"""
|
81 |
|
82 |
-
DEFAULT_TEMPLATE = """<|start_header_id|>system<|end_header_id|>
|
83 |
-
You are a helpful assistant. Use the following pieces of context to answer the question at the end.
|
84 |
-
If you don't know the answer, just say that you don't know, don't try to make up an answer.
|
85 |
-
|
86 |
-
Context:
|
87 |
-
{context}
|
88 |
-
|
89 |
-
<|eot_id|><|start_header_id|>user<|end_header_id|>
|
90 |
-
{question}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
|
91 |
-
"""
|
92 |
|
93 |
def __init__(
|
94 |
self,
|
@@ -104,10 +65,6 @@ Context:
|
|
104 |
self.last_context: Optional[str] = None
|
105 |
self.context = None
|
106 |
self.source_documents = 0
|
107 |
-
# self.prompt = PromptTemplate(
|
108 |
-
# template=self.DEFAULT_TEMPLATE,
|
109 |
-
# input_variables=["context", "question"]
|
110 |
-
# )
|
111 |
|
112 |
def _validate_file(self, file_path: Path) -> bool:
|
113 |
"""Validate if the file is of supported format and exists"""
|
@@ -191,48 +148,44 @@ Context:
|
|
191 |
retrieved_docs = retriever.get_relevant_documents(question)
|
192 |
context = self._format_context(retrieved_docs)
|
193 |
self.last_context = context
|
|
|
|
|
194 |
messages = [
|
195 |
{
|
196 |
"role":"system",
|
197 |
-
"content":f"""
|
198 |
-
You are a helpful assistant. Use the following pieces of context to answer the question at the end.
|
199 |
If you don't know the answer, just say that you don't know, don't try to make up an answer.
|
200 |
|
201 |
Context:
|
202 |
{context}
|
203 |
-
|
204 |
-
<|eot_id|><|start_header_id|>user<|end_header_id|>
|
205 |
-
{question}<|eot_id|><|start_header_id|>assistant<|end_header_id|>
|
206 |
"""
|
207 |
},
|
208 |
{
|
209 |
"role": "user",
|
210 |
-
"content":
|
211 |
}
|
212 |
]
|
213 |
-
self.context = context
|
214 |
-
self.source_documents = len(retrieved_docs)
|
215 |
-
# Generate response using LLM ###########
|
216 |
-
# response = self.llm.invoke(
|
217 |
-
# self.prompt.format(
|
218 |
-
# context=context,
|
219 |
-
# question=question
|
220 |
-
# )
|
221 |
-
# )
|
222 |
|
223 |
-
|
224 |
-
|
225 |
-
|
226 |
-
|
227 |
-
|
228 |
-
|
229 |
-
|
|
|
|
|
|
|
|
|
|
|
230 |
|
231 |
except Exception as e:
|
232 |
error_msg = f"Error during query processing: {str(e)}"
|
233 |
logger.error(error_msg)
|
234 |
-
|
|
|
235 |
|
|
|
236 |
def create_gradio_interface(rag_system: AdvancedRAGSystem) -> gr.Blocks:
|
237 |
"""Create an improved Gradio interface for the RAG system"""
|
238 |
|
@@ -274,14 +227,14 @@ def create_gradio_interface(rag_system: AdvancedRAGSystem) -> gr.Blocks:
|
|
274 |
chunk_size = gr.Slider(
|
275 |
minimum=100,
|
276 |
maximum=10000,
|
277 |
-
value=
|
278 |
step=100,
|
279 |
label="Chunk Size"
|
280 |
)
|
281 |
overlap = gr.Slider(
|
282 |
minimum=10,
|
283 |
maximum=5000,
|
284 |
-
value=
|
285 |
step=10,
|
286 |
label="Chunk Overlap"
|
287 |
)
|
@@ -315,40 +268,20 @@ def create_gradio_interface(rag_system: AdvancedRAGSystem) -> gr.Blocks:
|
|
315 |
)
|
316 |
|
317 |
query_button.click(
|
318 |
-
fn=
|
319 |
inputs=[question_input],
|
320 |
outputs=[answer_output],
|
321 |
api_name="stream_response",
|
322 |
-
|
323 |
-
)
|
324 |
-
|
325 |
-
query_button.click(
|
326 |
-
fn=update_history,
|
327 |
inputs=[question_input],
|
328 |
-
outputs=[
|
329 |
)
|
330 |
|
331 |
return demo
|
332 |
|
333 |
|
334 |
-
"""
|
335 |
-
For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
|
336 |
-
"""
|
337 |
-
# demo = gr.ChatInterface(
|
338 |
-
# respond,
|
339 |
-
# additional_inputs=[
|
340 |
-
# gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
|
341 |
-
# gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
|
342 |
-
# gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
|
343 |
-
# gr.Slider(
|
344 |
-
# minimum=0.1,
|
345 |
-
# maximum=1.0,
|
346 |
-
# value=0.95,
|
347 |
-
# step=0.05,
|
348 |
-
# label="Top-p (nucleus sampling)",
|
349 |
-
# ),
|
350 |
-
# ],
|
351 |
-
# )
|
352 |
rag_system = AdvancedRAGSystem(embed_model, client)
|
353 |
demo = create_gradio_interface(rag_system)
|
354 |
|
|
|
8 |
import gradio as gr
|
9 |
from langchain.text_splitter import RecursiveCharacterTextSplitter
|
10 |
from langchain_community.vectorstores import Chroma
|
|
|
11 |
from langchain.schema import BaseRetriever
|
12 |
from langchain.embeddings.base import Embeddings
|
13 |
from langchain.llms.base import BaseLanguageModel
|
14 |
import PyPDF2
|
15 |
+
from huggingface_hub import InferenceClient
|
16 |
# Install required packages
|
17 |
|
18 |
|
19 |
# Initialize models
|
20 |
import torch
|
|
|
21 |
from langchain_community.embeddings import HuggingFaceBgeEmbeddings
|
|
|
|
|
|
|
|
|
22 |
embed_model = HuggingFaceBgeEmbeddings(
|
23 |
model_name="all-MiniLM-L6-v2",#"dunzhang/stella_en_1.5B_v5",
|
24 |
model_kwargs={'device': 'cpu'},
|
|
|
26 |
)
|
27 |
|
28 |
model_name = "meta-llama/Llama-3.2-3B-Instruct"#"google/gemma-2-2b-it"#"prithivMLmods/Llama-3.2-3B-GGUF"
|
|
|
|
|
|
|
29 |
|
30 |
|
31 |
+
client = InferenceClient(model_name)
|
32 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
33 |
|
34 |
# Set up logging
|
35 |
logging.basicConfig(level=logging.INFO)
|
|
|
42 |
@dataclass
|
43 |
class RAGConfig:
|
44 |
"""Configuration for RAG system parameters"""
|
45 |
+
chunk_size: int = 100
|
46 |
+
chunk_overlap: int = 10
|
47 |
retriever_k: int = 3
|
48 |
persist_directory: str = "./chroma_db"
|
49 |
|
50 |
class AdvancedRAGSystem:
|
51 |
"""Advanced RAG System with improved error handling and type safety"""
|
52 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
53 |
|
54 |
def __init__(
|
55 |
self,
|
|
|
65 |
self.last_context: Optional[str] = None
|
66 |
self.context = None
|
67 |
self.source_documents = 0
|
|
|
|
|
|
|
|
|
68 |
|
69 |
def _validate_file(self, file_path: Path) -> bool:
|
70 |
"""Validate if the file is of supported format and exists"""
|
|
|
148 |
retrieved_docs = retriever.get_relevant_documents(question)
|
149 |
context = self._format_context(retrieved_docs)
|
150 |
self.last_context = context
|
151 |
+
self.context = context
|
152 |
+
self.source_documents = len(retrieved_docs)
|
153 |
messages = [
|
154 |
{
|
155 |
"role":"system",
|
156 |
+
"content":f"""You are a helpful assistant. Use the following pieces of context to answer the question at the end.
|
|
|
157 |
If you don't know the answer, just say that you don't know, don't try to make up an answer.
|
158 |
|
159 |
Context:
|
160 |
{context}
|
|
|
|
|
|
|
161 |
"""
|
162 |
},
|
163 |
{
|
164 |
"role": "user",
|
165 |
+
"content": question
|
166 |
}
|
167 |
]
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
168 |
|
169 |
+
response_text = ""
|
170 |
+
for chunk in self.llm.chat.completions.create(
|
171 |
+
model=model_name,
|
172 |
+
messages=messages,
|
173 |
+
max_tokens=500,
|
174 |
+
stream=True
|
175 |
+
):
|
176 |
+
if hasattr(chunk.choices[0].delta, 'content'):
|
177 |
+
content = chunk.choices[0].delta.content
|
178 |
+
if content is not None:
|
179 |
+
response_text += content
|
180 |
+
yield response_text
|
181 |
|
182 |
except Exception as e:
|
183 |
error_msg = f"Error during query processing: {str(e)}"
|
184 |
logger.error(error_msg)
|
185 |
+
yield error_msg
|
186 |
+
|
187 |
|
188 |
+
|
189 |
def create_gradio_interface(rag_system: AdvancedRAGSystem) -> gr.Blocks:
|
190 |
"""Create an improved Gradio interface for the RAG system"""
|
191 |
|
|
|
227 |
chunk_size = gr.Slider(
|
228 |
minimum=100,
|
229 |
maximum=10000,
|
230 |
+
value=100,
|
231 |
step=100,
|
232 |
label="Chunk Size"
|
233 |
)
|
234 |
overlap = gr.Slider(
|
235 |
minimum=10,
|
236 |
maximum=5000,
|
237 |
+
value=10,
|
238 |
step=10,
|
239 |
label="Chunk Overlap"
|
240 |
)
|
|
|
268 |
)
|
269 |
|
270 |
query_button.click(
|
271 |
+
fn=query_streaming,
|
272 |
inputs=[question_input],
|
273 |
outputs=[answer_output],
|
274 |
api_name="stream_response",
|
275 |
+
queue=False
|
276 |
+
).then(
|
277 |
+
fn=update_context,
|
|
|
|
|
278 |
inputs=[question_input],
|
279 |
+
outputs=[context_output]
|
280 |
)
|
281 |
|
282 |
return demo
|
283 |
|
284 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
285 |
rag_system = AdvancedRAGSystem(embed_model, client)
|
286 |
demo = create_gradio_interface(rag_system)
|
287 |
|