ShadowHunter222 commited on
Commit
bc39393
·
verified ·
1 Parent(s): f2c5b0b

Update Dockerfile

Browse files
Files changed (1) hide show
  1. Dockerfile +4 -8
Dockerfile CHANGED
@@ -2,13 +2,9 @@ FROM ghcr.io/ggml-org/llama.cpp:server
2
 
3
  RUN apt update && apt install wget -y && rm -rf /var/lib/apt/lists/*
4
 
5
- # THE MAIN MODEL: LiquidAI LFM 2.5 VL 1.6B (Q4_0)
6
- # Note: Q4_K_M is not officially provided in GGUF for this model, using Q4_0 which is the closest 4-bit quant.
7
- RUN wget "https://huggingface.co/LiquidAI/LFM2.5-VL-1.6B-GGUF/resolve/main/LFM2.5-VL-1.6B-Q4_0.gguf" -O /model.gguf
8
 
9
- # THE MULTIMODAL PROJECTOR (Vision)
10
- RUN wget "https://huggingface.co/LiquidAI/LFM2.5-VL-1.6B-GGUF/resolve/main/mmproj-LFM2.5-VL-1.6b-F16.gguf" -O /mmproj.gguf
11
-
12
- # Run the server with multimodal projector for vision support
13
  ENTRYPOINT []
14
- CMD /app/llama-server -m /model.gguf --mmproj /mmproj.gguf --port 7860 --host 0.0.0.0 -t 2 --mlock -c 4096 --log-disable > /dev/null 2>&1
 
2
 
3
  RUN apt update && apt install wget -y && rm -rf /var/lib/apt/lists/*
4
 
5
+ # THE MAIN MODEL: LiquidAI LFM3-1.7B (Q4_K_M)
6
+ RUN wget "https://huggingface.co/LiquidAI/LFM3-1.7B-GGUF/resolve/main/LFM3-1.7B-Q4_K_M.gguf" -O /model.gguf
 
7
 
8
+ # Run the server with official LFM3 sampling parameters
 
 
 
9
  ENTRYPOINT []
10
+ CMD /app/llama-server -m /model.gguf --port 7860 --host 0.0.0.0 -t 2 -c 4096 --temp 0.1 --top-k 50 --repeat-penalty 1.1 --log-disable > /dev/null 2>&1