hailin
/
embed-bge-m3
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

.

This commit is contained in:
hailin 2025-08-10 23:55:30 +08:00
parent 640a0a7712
commit 72a6daa3ac
1 changed files with 3 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
app/main.py
View File

@ -25,9 +25,9 @@ from FlagEmbedding import BGEM3FlagModel
# Config # Config
# -----------------------------------------------------------------------------# # -----------------------------------------------------------------------------#
MODEL_PATH = "model/bge-m3" # 按需改成你的权重路径 MODEL_PATH = "model/bge-m3" # 按需改成你的权重路径
MODEL_VRAM_MB = int(os.getenv("MODEL_VRAM_MB", "8000")) # bge-m3-large fp32 ≈ 8 GiB MODEL_VRAM_MB = int(os.getenv("MODEL_VRAM_MB", "4800")) # bge-m3-large fp16=2.4 fp32 ≈ 4.8 GiB
POST_LOAD_GAP_MB = 192 POST_LOAD_GAP_MB = 200
SAFE_MIN_FREE_MB = MODEL_VRAM_MB + POST_LOAD_GAP_MB # == 8192 MB SAFE_MIN_FREE_MB = MODEL_VRAM_MB + POST_LOAD_GAP_MB # == 5000 MB
# 请求批次与单条最大长度上限（防御异常大 payload） # 请求批次与单条最大长度上限（防御异常大 payload）
MAX_BATCH = int(os.getenv("MAX_BATCH", "1024")) MAX_BATCH = int(os.getenv("MAX_BATCH", "1024"))