rikunarita commited on
Commit
e68d948
·
verified ·
1 Parent(s): 2d1f03c

Create Dockerfile

Browse files
Files changed (1) hide show
  1. Dockerfile +25 -0
Dockerfile ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ FROM ghcr.io/ggml-org/llama.cpp:full
2
+
3
+ WORKDIR /app
4
+
5
+ # パッケージのインストール
6
+ RUN apt update && apt install -y python3-pip
7
+ RUN pip install -U huggingface_hub
8
+
9
+ # モデルとmmprojの両方をHugging Faceからダウンロード
10
+ RUN python3 -c 'from huggingface_hub import hf_hub_download; \
11
+ repo="HauhauCS/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive"; \
12
+ hf_hub_download(repo_id=repo, filename="Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q6_K_P.gguf", local_dir="/app"); \
13
+ hf_hub_download(repo_id=repo, filename="mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.gguf", local_dir="/app")'
14
+
15
+ # サーバーの設定(mmprojを追加)
16
+ CMD ["--server", \
17
+ "-m", "/app/Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-Q6_K_P.gguf", \
18
+ "--mmproj", "/app/mmproj-Qwen3.6-35B-A3B-Uncensored-HauhauCS-Aggressive-f16.gguf", \
19
+ "--host", "0.0.0.0", \
20
+ "--port", "7860", \
21
+ "-t", "2", \
22
+ "--cache-type-k", "q8_0", \
23
+ "--cache-type-v", "iq4_nl", \
24
+ "-c", "64000", \
25
+ "-n", "38912"]