Spaces:
Paused
Paused
modify load timing of model
Browse files
app.py
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
-
|
2 |
import gradio as gr
|
3 |
from langchain.chains import RetrievalQA
|
4 |
from langchain.embeddings import OpenAIEmbeddings
|
@@ -16,17 +16,29 @@ from qdrant_client import QdrantClient
|
|
16 |
from config import DB_CONFIG, DB_E5_CONFIG
|
17 |
|
18 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
19 |
def _get_config_and_embeddings(collection_name: str | None) -> tuple:
|
20 |
if collection_name is None or collection_name == "E5":
|
21 |
db_config = DB_E5_CONFIG
|
22 |
-
|
23 |
-
model_kwargs = {"device": "cpu"}
|
24 |
-
encode_kwargs = {"normalize_embeddings": False}
|
25 |
-
embeddings = HuggingFaceEmbeddings(
|
26 |
-
model_name=model_name,
|
27 |
-
model_kwargs=model_kwargs,
|
28 |
-
encode_kwargs=encode_kwargs,
|
29 |
-
)
|
30 |
elif collection_name == "OpenAI":
|
31 |
db_config = DB_CONFIG
|
32 |
embeddings = OpenAIEmbeddings()
|
@@ -36,18 +48,10 @@ def _get_config_and_embeddings(collection_name: str | None) -> tuple:
|
|
36 |
|
37 |
|
38 |
def _get_rinna_llm(temperature: float):
|
39 |
-
model = "rinna/bilingual-gpt-neox-4b-instruction-ppo"
|
40 |
-
tokenizer = AutoTokenizer.from_pretrained(model, use_fast=False)
|
41 |
-
model = AutoModelForCausalLM.from_pretrained(
|
42 |
-
model,
|
43 |
-
load_in_8bit=True,
|
44 |
-
torch_dtype=torch.float16,
|
45 |
-
device_map="auto",
|
46 |
-
)
|
47 |
pipe = pipeline(
|
48 |
"text-generation",
|
49 |
-
model=
|
50 |
-
tokenizer=
|
51 |
max_new_tokens=1024,
|
52 |
temperature=temperature,
|
53 |
)
|
@@ -139,6 +143,7 @@ def get_related_url(metadata):
|
|
139 |
def main(
|
140 |
query: str, collection_name: str, model_name: str, option: str, temperature: float
|
141 |
):
|
|
|
142 |
qa = get_retrieval_qa(collection_name, model_name, temperature, option)
|
143 |
try:
|
144 |
result = qa(query)
|
@@ -146,7 +151,8 @@ def main(
|
|
146 |
return "回答が見つかりませんでした。別な質問をしてみてください", str(e)
|
147 |
else:
|
148 |
metadata = [s.metadata for s in result["source_documents"]]
|
149 |
-
|
|
|
150 |
|
151 |
return result["result"], html
|
152 |
|
|
|
1 |
+
from time import time
|
2 |
import gradio as gr
|
3 |
from langchain.chains import RetrievalQA
|
4 |
from langchain.embeddings import OpenAIEmbeddings
|
|
|
16 |
from config import DB_CONFIG, DB_E5_CONFIG
|
17 |
|
18 |
|
19 |
+
E5_MODEL_NAME = "intfloat/multilingual-e5-large"
|
20 |
+
E5_MODEL_KWARGS = {"device": "cuda:0" if torch.cuda.is_available() else "cpu"}
|
21 |
+
E5_ENCODE_KWARGS = {"normalize_embeddings": False}
|
22 |
+
E5_EMBEDDINGS = HuggingFaceEmbeddings(
|
23 |
+
model_name=E5_MODEL_NAME,
|
24 |
+
model_kwargs=E5_MODEL_KWARGS,
|
25 |
+
encode_kwargs=E5_ENCODE_KWARGS,
|
26 |
+
)
|
27 |
+
|
28 |
+
RINNA_MODEL_NAME = "rinna/bilingual-gpt-neox-4b-instruction-ppo"
|
29 |
+
RINNA_TOKENIZER = AutoTokenizer.from_pretrained(RINNA_MODEL_NAME, use_fast=False)
|
30 |
+
RINNA_MODEL = AutoModelForCausalLM.from_pretrained(
|
31 |
+
RINNA_MODEL_NAME,
|
32 |
+
load_in_8bit=True,
|
33 |
+
torch_dtype=torch.float16,
|
34 |
+
device_map="auto",
|
35 |
+
)
|
36 |
+
|
37 |
+
|
38 |
def _get_config_and_embeddings(collection_name: str | None) -> tuple:
|
39 |
if collection_name is None or collection_name == "E5":
|
40 |
db_config = DB_E5_CONFIG
|
41 |
+
embeddings = E5_EMBEDDINGS
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
42 |
elif collection_name == "OpenAI":
|
43 |
db_config = DB_CONFIG
|
44 |
embeddings = OpenAIEmbeddings()
|
|
|
48 |
|
49 |
|
50 |
def _get_rinna_llm(temperature: float):
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
51 |
pipe = pipeline(
|
52 |
"text-generation",
|
53 |
+
model=RINNA_MODEL,
|
54 |
+
tokenizer=RINNA_TOKENIZER,
|
55 |
max_new_tokens=1024,
|
56 |
temperature=temperature,
|
57 |
)
|
|
|
143 |
def main(
|
144 |
query: str, collection_name: str, model_name: str, option: str, temperature: float
|
145 |
):
|
146 |
+
now = time()
|
147 |
qa = get_retrieval_qa(collection_name, model_name, temperature, option)
|
148 |
try:
|
149 |
result = qa(query)
|
|
|
151 |
return "回答が見つかりませんでした。別な質問をしてみてください", str(e)
|
152 |
else:
|
153 |
metadata = [s.metadata for s in result["source_documents"]]
|
154 |
+
sec_html = f"<p>実行時間: {(time() - now):.2f}秒</p>"
|
155 |
+
html = "<div>" + sec_html + "\n".join(get_related_url(metadata)) + "</div>"
|
156 |
|
157 |
return result["result"], html
|
158 |
|