Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -10,16 +10,14 @@ from accelerate import Accelerator
|
|
10 |
# νκ²½ λ³μμμ Hugging Face API ν€ λ‘λ
|
11 |
hf_api_key = os.getenv('HF_API_KEY')
|
12 |
|
13 |
-
# λͺ¨λΈ ID μ€μ
|
14 |
model_id = "microsoft/phi-2"
|
15 |
-
|
16 |
-
# μ¬μ©μ μ μ μ½λ μ€ν νμ©κ³Ό ν¨κ» λͺ¨λΈ λ° ν ν¬λμ΄μ λ‘λ
|
17 |
tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_api_key, trust_remote_code=True)
|
18 |
model = AutoModelForCausalLM.from_pretrained(
|
19 |
model_id,
|
20 |
token=hf_api_key,
|
21 |
-
trust_remote_code=True,
|
22 |
-
torch_dtype=torch.float32
|
23 |
)
|
24 |
|
25 |
# Accelerator μ€μ
|
@@ -32,11 +30,10 @@ dataset = load_dataset("not-lain/wikipedia", revision="embedded")
|
|
32 |
data = dataset["train"]
|
33 |
data = data.add_faiss_index("embeddings")
|
34 |
|
35 |
-
# κ²μ, ν둬ννΈ ν¬λ§·ν
, μλ΅ μμ± ν¨μ
|
36 |
def search(query: str, k: int = 3):
|
37 |
embedded_query = ST.encode(query)
|
38 |
scores, retrieved_examples = data.get_nearest_examples("embeddings", embedded_query, k=k)
|
39 |
-
return scores,
|
40 |
|
41 |
def format_prompt(prompt, retrieved_documents, k):
|
42 |
PROMPT = f"Question:{prompt}\nContext:"
|
@@ -55,16 +52,15 @@ def rag_chatbot_interface(prompt: str, k: int = 2):
|
|
55 |
formatted_prompt = format_prompt(prompt, retrieved_documents, k)
|
56 |
return generate(formatted_prompt)
|
57 |
|
58 |
-
# μμ€ν
ν둬ννΈ
|
59 |
SYS_PROMPT = "You are an assistant for answering questions. Provide a conversational answer."
|
60 |
|
61 |
-
# Gradio μΈν°νμ΄μ€
|
62 |
iface = gr.Interface(
|
63 |
fn=rag_chatbot_interface,
|
64 |
inputs="text",
|
65 |
outputs="text",
|
66 |
title="Retrieval-Augmented Generation Chatbot",
|
67 |
-
description="This chatbot provides more accurate answers by searching relevant documents and generating responses."
|
|
|
68 |
)
|
69 |
|
70 |
iface.launch()
|
|
|
10 |
# νκ²½ λ³μμμ Hugging Face API ν€ λ‘λ
|
11 |
hf_api_key = os.getenv('HF_API_KEY')
|
12 |
|
13 |
+
# λͺ¨λΈ ID λ° ν ν¬λμ΄μ μ€μ
|
14 |
model_id = "microsoft/phi-2"
|
|
|
|
|
15 |
tokenizer = AutoTokenizer.from_pretrained(model_id, token=hf_api_key, trust_remote_code=True)
|
16 |
model = AutoModelForCausalLM.from_pretrained(
|
17 |
model_id,
|
18 |
token=hf_api_key,
|
19 |
+
trust_remote_code=True,
|
20 |
+
torch_dtype=torch.float32
|
21 |
)
|
22 |
|
23 |
# Accelerator μ€μ
|
|
|
30 |
data = dataset["train"]
|
31 |
data = data.add_faiss_index("embeddings")
|
32 |
|
|
|
33 |
def search(query: str, k: int = 3):
|
34 |
embedded_query = ST.encode(query)
|
35 |
scores, retrieved_examples = data.get_nearest_examples("embeddings", embedded_query, k=k)
|
36 |
+
return scores, retrieved_examples
|
37 |
|
38 |
def format_prompt(prompt, retrieved_documents, k):
|
39 |
PROMPT = f"Question:{prompt}\nContext:"
|
|
|
52 |
formatted_prompt = format_prompt(prompt, retrieved_documents, k)
|
53 |
return generate(formatted_prompt)
|
54 |
|
|
|
55 |
SYS_PROMPT = "You are an assistant for answering questions. Provide a conversational answer."
|
56 |
|
|
|
57 |
iface = gr.Interface(
|
58 |
fn=rag_chatbot_interface,
|
59 |
inputs="text",
|
60 |
outputs="text",
|
61 |
title="Retrieval-Augmented Generation Chatbot",
|
62 |
+
description="This chatbot provides more accurate answers by searching relevant documents and generating responses.",
|
63 |
+
share=True # κ³΅κ° λ§ν¬ μμ±
|
64 |
)
|
65 |
|
66 |
iface.launch()
|