Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
@@ -25,22 +25,13 @@ torch_device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
25 |
print("Running on device:", torch_device)
|
26 |
print("CPU threads:", torch.get_num_threads())
|
27 |
|
28 |
-
loader = PyPDFLoader("total.pdf")
|
29 |
-
pages = loader.load()
|
30 |
-
|
31 |
-
# λ°μ΄ν°λ₯Ό λΆλ¬μμ ν
μ€νΈλ₯Ό μΌμ ν μλ‘ λλκ³ κ΅¬λΆμλ‘ μ°κ²°νλ μμ
|
32 |
-
text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=0)
|
33 |
-
texts = text_splitter.split_documents(pages)
|
34 |
-
|
35 |
-
print(f"λ¬Έμμ {len(texts)}κ°μ λ¬Έμλ₯Ό κ°μ§κ³ μμ΅λλ€.")
|
36 |
-
|
37 |
hf_hub_download(repo_id='StarFox7/Llama-2-ko-7B-chat-gguf', filename='Llama-2-ko-7B-chat-gguf-q4_0.bin', local_dir='./')
|
38 |
|
39 |
llm = LlamaCpp(
|
40 |
model_path='Llama-2-ko-7B-chat-gguf-q4_0.bin',
|
41 |
temperature=0.5,
|
42 |
top_p=0.9,
|
43 |
-
max_tokens=
|
44 |
verbose=True,
|
45 |
n_ctx=2048,
|
46 |
n_gpu_layers=-1,
|
@@ -50,16 +41,8 @@ llm = LlamaCpp(
|
|
50 |
# μλ² λ© λͺ¨λΈ λ‘λ
|
51 |
embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")
|
52 |
|
53 |
-
# λ¬Έμμ μλ ν
μ€νΈλ₯Ό μλ² λ©νκ³ FAISS μ μΈλ±μ€λ₯Ό ꡬμΆν¨
|
54 |
-
index = FAISS.from_documents(
|
55 |
-
documents=texts,
|
56 |
-
embedding=embeddings,
|
57 |
-
)
|
58 |
-
|
59 |
-
# faiss_db λ‘ λ‘컬μ μ μ₯νκΈ°
|
60 |
-
index.save_local("faiss_db")
|
61 |
# faiss_db λ‘ λ‘컬μ λ‘λνκΈ°
|
62 |
-
docsearch = FAISS.load_local("
|
63 |
|
64 |
embeddings_filter = EmbeddingsFilter(
|
65 |
embeddings=embeddings,
|
@@ -118,13 +101,11 @@ def gen(x, id, customer_data):
|
|
118 |
|
119 |
### λͺ
λ Ήμ΄:
|
120 |
λ€μ μ§μΉ¨μ μ°Έκ³ νμ¬ μλ΄μμΌλ‘μ κ³ κ°μκ² νμν μλ΅μ μ 곡νμΈμ.
|
121 |
-
|
122 |
[μ§μΉ¨]
|
123 |
1.κ³ κ°μ κ°μ
μ 보λ₯Ό κΌ νμΈνμ¬ κ³ κ°μ΄ κ°μ
ν 보νμ λν λ΄μ©λ§ μ 곡νμΈμ.
|
124 |
2.κ³ κ°μ΄ κ°μ
ν 보νμ΄λΌλ©΄ κ³ κ°μ μ§λ¬Έμ λν΄ μ μ ν λ΅λ³νμΈμ.
|
125 |
3.κ³ κ°μ΄ κ°μ
νμ§ μμ 보νμ 보μμ κ΄ν μ§λ¬Έμ κ΄λ ¨ 보νμ μκ°νλ©° 보μμ΄ λΆκ°λ₯νλ€λ μ μ μλ΄νμΈμ.
|
126 |
4.κ³ κ°μ΄ κ°μ
νμ§ μμ 보νμ κ°μ
μ΄ νμνλ€κ³ 보νλͺ
μ νμ€νκ² μΈκΈνμΈμ.
|
127 |
-
|
128 |
λ€μ μ
λ ₯μ μ£Όμ΄μ§λ κ³ κ°μ 보ν κ°μ
μ 보μ μλ΄ κΈ°λ‘μ λ³΄κ³ κ³ κ°μκ² λμλλ μ 보λ₯Ό μ 곡νμΈμ. μ°¨κ·Όμ°¨κ·Ό μκ°νμ¬ λ΅λ³νμΈμ. λΉμ μ μ ν μ μμ΅λλ€.
|
129 |
|
130 |
### μ
λ ₯:
|
|
|
25 |
print("Running on device:", torch_device)
|
26 |
print("CPU threads:", torch.get_num_threads())
|
27 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
28 |
hf_hub_download(repo_id='StarFox7/Llama-2-ko-7B-chat-gguf', filename='Llama-2-ko-7B-chat-gguf-q4_0.bin', local_dir='./')
|
29 |
|
30 |
llm = LlamaCpp(
|
31 |
model_path='Llama-2-ko-7B-chat-gguf-q4_0.bin',
|
32 |
temperature=0.5,
|
33 |
top_p=0.9,
|
34 |
+
max_tokens=80,
|
35 |
verbose=True,
|
36 |
n_ctx=2048,
|
37 |
n_gpu_layers=-1,
|
|
|
41 |
# μλ² λ© λͺ¨λΈ λ‘λ
|
42 |
embeddings = HuggingFaceEmbeddings(model_name="intfloat/multilingual-e5-large")
|
43 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
44 |
# faiss_db λ‘ λ‘컬μ λ‘λνκΈ°
|
45 |
+
docsearch = FAISS.load_local("", embeddings)
|
46 |
|
47 |
embeddings_filter = EmbeddingsFilter(
|
48 |
embeddings=embeddings,
|
|
|
101 |
|
102 |
### λͺ
λ Ήμ΄:
|
103 |
λ€μ μ§μΉ¨μ μ°Έκ³ νμ¬ μλ΄μμΌλ‘μ κ³ κ°μκ² νμν μλ΅μ μ 곡νμΈμ.
|
|
|
104 |
[μ§μΉ¨]
|
105 |
1.κ³ κ°μ κ°μ
μ 보λ₯Ό κΌ νμΈνμ¬ κ³ κ°μ΄ κ°μ
ν 보νμ λν λ΄μ©λ§ μ 곡νμΈμ.
|
106 |
2.κ³ κ°μ΄ κ°μ
ν 보νμ΄λΌλ©΄ κ³ κ°μ μ§λ¬Έμ λν΄ μ μ ν λ΅λ³νμΈμ.
|
107 |
3.κ³ κ°μ΄ κ°μ
νμ§ μμ 보νμ 보μμ κ΄ν μ§λ¬Έμ κ΄λ ¨ 보νμ μκ°νλ©° 보μμ΄ λΆκ°λ₯νλ€λ μ μ μλ΄νμΈμ.
|
108 |
4.κ³ κ°μ΄ κ°μ
νμ§ μμ 보νμ κ°μ
μ΄ νμνλ€κ³ 보νλͺ
μ νμ€νκ² μΈκΈνμΈμ.
|
|
|
109 |
λ€μ μ
λ ₯μ μ£Όμ΄μ§λ κ³ κ°μ 보ν κ°μ
μ 보μ μλ΄ κΈ°λ‘μ λ³΄κ³ κ³ κ°μκ² λμλλ μ 보λ₯Ό μ 곡νμΈμ. μ°¨κ·Όμ°¨κ·Ό μκ°νμ¬ λ΅λ³νμΈμ. λΉμ μ μ ν μ μμ΅λλ€.
|
110 |
|
111 |
### μ
λ ₯:
|