ffreemt
commited on
Commit
•
0e167ee
1
Parent(s):
7f5a78c
Update os.environ["HF_HOME"]="/tmp/cache"
Browse files- Dockerfile +0 -1
- m3_server.py +3 -0
Dockerfile
CHANGED
@@ -22,5 +22,4 @@ RUN pip install --no-cache-dir --upgrade pip && \
|
|
22 |
# CMD ["python", "app.py"]
|
23 |
|
24 |
# CMD ["sh", "start-m3-server.sh"]
|
25 |
-
|
26 |
CMD ["python", "m3_server.py"]
|
|
|
22 |
# CMD ["python", "app.py"]
|
23 |
|
24 |
# CMD ["sh", "start-m3-server.sh"]
|
|
|
25 |
CMD ["python", "m3_server.py"]
|
m3_server.py
CHANGED
@@ -1,3 +1,4 @@
|
|
|
|
1 |
import asyncio
|
2 |
import time
|
3 |
from concurrent.futures import ThreadPoolExecutor
|
@@ -10,6 +11,8 @@ from FlagEmbedding import BGEM3FlagModel
|
|
10 |
from pydantic import BaseModel
|
11 |
from starlette.status import HTTP_504_GATEWAY_TIMEOUT
|
12 |
|
|
|
|
|
13 |
batch_size = 2 # gpu batch_size in order of your available vram
|
14 |
max_request = 10 # max request for future improvements on api calls / gpu batches (for now is pretty basic)
|
15 |
max_length = 5000 # max context length for embeddings and passages in re-ranker
|
|
|
1 |
+
impor os
|
2 |
import asyncio
|
3 |
import time
|
4 |
from concurrent.futures import ThreadPoolExecutor
|
|
|
11 |
from pydantic import BaseModel
|
12 |
from starlette.status import HTTP_504_GATEWAY_TIMEOUT
|
13 |
|
14 |
+
os.environ["HF_HOME"] = "/tmp/cache"
|
15 |
+
|
16 |
batch_size = 2 # gpu batch_size in order of your available vram
|
17 |
max_request = 10 # max request for future improvements on api calls / gpu batches (for now is pretty basic)
|
18 |
max_length = 5000 # max context length for embeddings and passages in re-ranker
|