Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -1,17 +1,18 @@
|
|
1 |
-
import gradio as gr
|
2 |
import torch
|
3 |
from transformers import AutoTokenizer, pipeline
|
4 |
from sentence_transformers import SentenceTransformer
|
5 |
import faiss
|
6 |
import numpy as np
|
|
|
|
|
7 |
|
8 |
# Configuration
|
9 |
class Config:
|
10 |
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
|
11 |
embedding_model = "all-MiniLM-L6-v2"
|
12 |
-
vector_dim = 384
|
13 |
-
top_k = 3
|
14 |
-
chunk_size = 256
|
15 |
|
16 |
# Vector Database
|
17 |
class VectorDB:
|
@@ -27,7 +28,7 @@ class VectorDB:
|
|
27 |
self.index.add(embedding)
|
28 |
self.texts.append(text)
|
29 |
|
30 |
-
def search(self, query: str):
|
31 |
if self.index.ntotal == 0:
|
32 |
return []
|
33 |
query_embedding = self.embedding_model.encode([query])[0]
|
@@ -42,7 +43,7 @@ class TinyChatModel:
|
|
42 |
self.tokenizer = AutoTokenizer.from_pretrained(Config.model_name)
|
43 |
self.pipe = pipeline("text-generation", model=Config.model_name, torch_dtype=torch.bfloat16, device_map="auto")
|
44 |
|
45 |
-
def generate_response(self, message: str, context: str = ""):
|
46 |
messages = [{"role": "user", "content": message}]
|
47 |
if context:
|
48 |
messages.insert(0, {"role": "system", "content": f"Context:\n{context}"})
|
@@ -54,32 +55,35 @@ class TinyChatModel:
|
|
54 |
vector_db = VectorDB()
|
55 |
chat_model = TinyChatModel()
|
56 |
|
57 |
-
|
58 |
-
|
59 |
-
|
|
|
|
|
|
|
|
|
|
|
60 |
vector_db.add_text(f"User: {user_input}\nAssistant: {response}")
|
|
|
61 |
return response
|
62 |
|
63 |
-
|
64 |
-
|
65 |
-
|
|
|
66 |
|
67 |
-
# Gradio UI
|
68 |
-
|
69 |
-
|
70 |
-
gr.Markdown("# 🦙 TinyChat - AI Chatbot")
|
71 |
-
with gr.Row():
|
72 |
-
chatbot = gr.Chatbot()
|
73 |
with gr.Row():
|
74 |
-
|
75 |
-
|
76 |
-
|
77 |
-
|
78 |
-
|
79 |
-
|
80 |
-
|
81 |
-
add_text_btn.click(add_text_interface, inputs=add_text_input, outputs=gr.Textbox())
|
82 |
|
83 |
-
#
|
84 |
if __name__ == "__main__":
|
85 |
-
demo.launch()
|
|
|
|
|
1 |
import torch
|
2 |
from transformers import AutoTokenizer, pipeline
|
3 |
from sentence_transformers import SentenceTransformer
|
4 |
import faiss
|
5 |
import numpy as np
|
6 |
+
import gradio as gr
|
7 |
+
from typing import List
|
8 |
|
9 |
# Configuration
|
10 |
class Config:
|
11 |
model_name = "TinyLlama/TinyLlama-1.1B-Chat-v1.0"
|
12 |
embedding_model = "all-MiniLM-L6-v2"
|
13 |
+
vector_dim = 384 # Sentence Transformer embedding dimension
|
14 |
+
top_k = 3 # Retrieve top 3 relevant chunks
|
15 |
+
chunk_size = 256 # Text chunk size
|
16 |
|
17 |
# Vector Database
|
18 |
class VectorDB:
|
|
|
28 |
self.index.add(embedding)
|
29 |
self.texts.append(text)
|
30 |
|
31 |
+
def search(self, query: str) -> List[str]:
|
32 |
if self.index.ntotal == 0:
|
33 |
return []
|
34 |
query_embedding = self.embedding_model.encode([query])[0]
|
|
|
43 |
self.tokenizer = AutoTokenizer.from_pretrained(Config.model_name)
|
44 |
self.pipe = pipeline("text-generation", model=Config.model_name, torch_dtype=torch.bfloat16, device_map="auto")
|
45 |
|
46 |
+
def generate_response(self, message: str, context: str = "") -> str:
|
47 |
messages = [{"role": "user", "content": message}]
|
48 |
if context:
|
49 |
messages.insert(0, {"role": "system", "content": f"Context:\n{context}"})
|
|
|
55 |
vector_db = VectorDB()
|
56 |
chat_model = TinyChatModel()
|
57 |
|
58 |
+
# Function to handle context addition and chat
|
59 |
+
def chat_function(user_input: str, context: str = ""):
|
60 |
+
if context:
|
61 |
+
vector_db.add_text(context)
|
62 |
+
|
63 |
+
# Search relevant context
|
64 |
+
context_text = "\n".join(vector_db.search(user_input))
|
65 |
+
response = chat_model.generate_response(user_input, context_text)
|
66 |
vector_db.add_text(f"User: {user_input}\nAssistant: {response}")
|
67 |
+
|
68 |
return response
|
69 |
|
70 |
+
# Gradio Interface
|
71 |
+
def gradio_interface(user_input: str, context: str = ""):
|
72 |
+
response = chat_function(user_input, context)
|
73 |
+
return response
|
74 |
|
75 |
+
# Create Gradio UI
|
76 |
+
with gr.Blocks() as demo:
|
77 |
+
gr.Markdown("# TinyChat: A Conversational AI")
|
|
|
|
|
|
|
78 |
with gr.Row():
|
79 |
+
with gr.Column():
|
80 |
+
user_input = gr.Textbox(label="User Input", placeholder="Ask anything...")
|
81 |
+
context_input = gr.Textbox(label="Optional Context", placeholder="Paste context here (optional)", lines=3)
|
82 |
+
submit_button = gr.Button("Send")
|
83 |
+
output = gr.Textbox(label="Response", placeholder="Assistant's reply will appear here...")
|
84 |
+
|
85 |
+
submit_button.click(fn=gradio_interface, inputs=[user_input, context_input], outputs=output)
|
|
|
86 |
|
87 |
+
# Run the Gradio app
|
88 |
if __name__ == "__main__":
|
89 |
+
demo.launch(share=True)
|