「笔记」 Docker 部署 BGE-M3 Embedder 模型
模型地址
https://huggingface.co/bbvch-ai/bge-m3-GGUF/tree/main
下载 bge-m3-q4_k_m.gguf 放到 models 文件夹
Compose
文档: https://github.com/ggerganov/llama.cpp/blob/master/examples/server/README.md
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
| services:
embedder-server:
runtime: nvidia
image: ghcr.io/ggerganov/llama.cpp:server-cuda
container_name: librechat-embedder
restart: unless-stopped
networks:
- traefik
command: ["--keep", "-1"]
environment:
TZ: "Asia/Shanghai"
LLAMA_ARG_HOST: "0.0.0.0"
LLAMA_ARG_PORT: "8080"
LLAMA_ARG_NO_WEBUI: true
LLAMA_ARG_MODEL: "/models/bge-m3-q4_k_m.gguf"
LLAMA_ARG_CTX_SIZE: 0
LLAMA_ARG_N_PARALLEL: 4
LLAMA_ARG_N_GPU_LAYERS: 99
LLAMA_ARG_ALIAS: "bge-m3-q4"
LLAMA_ARG_EMBEDDINGS: true
LLAMA_API_KEY: "sk-llama"
volumes:
- ./models:/models
|
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
| services:
embedder-server:
image: ghcr.io/ggerganov/llama.cpp:server
container_name: librechat-embedder
restart: unless-stopped
networks:
- traefik
command: ["--keep", "-1"]
environment:
TZ: "Asia/Shanghai"
LLAMA_ARG_HOST: "0.0.0.0"
LLAMA_ARG_PORT: "8080"
LLAMA_ARG_NO_WEBUI: true
LLAMA_ARG_MODEL: "/models/bge-m3-q4_k_m.gguf"
LLAMA_ARG_CTX_SIZE: 0
LLAMA_ARG_N_PARALLEL: 4
LLAMA_ARG_N_GPU_LAYERS: 99
LLAMA_ARG_ALIAS: "bge-m3-q4"
LLAMA_ARG_EMBEDDINGS: true
LLAMA_API_KEY: "sk-llama"
volumes:
- ./models:/models
|