Update Dockerfile

Dockerfile

@@ -4,10 +4,21 @@ FROM ghcr.io/ggml-org/llama.cpp:server
 RUN apt update && apt install wget -y && rm -rf /var/lib/apt/lists/*
 
 
-# 1. THE MAIN MODEL: LFM2.5-2.6B in Q4_K_M for balanced speed and performance on CPU
+# 1. THE MAIN MODEL: LFM2.5-2.6B in Q4_K_M — best speed/quality balance for CPU
 RUN wget "https://huggingface.co/LiquidAI/LFM2.5-2.6B-GGUF/resolve/main/LFM2.5-2.6B-Q4_K_M.gguf" -O /model.gguf
 
 
 # Clear the entrypoint and use shell form CMD to silence all logs
 ENTRYPOINT []
-CMD /app/llama-server -m /model.gguf --port 7860 --host 0.0.0.0 -t 2 --mlock --no-mmap -c 8192 --batch-size 512 --ubatch-size 512 --log-disable > /dev/null 2>&1
+CMD /app/llama-server \
+    -m /model.gguf \
+    --port 7860 --host 0.0.0.0 \
+    -t 2 --threads-batch 2 \
+    -c 4096 \
+    --parallel 1 \
+    --batch-size 256 --ubatch-size 256 \
+    --flash-attn \
+    --cache-type-k q8_0 --cache-type-v q8_0 \
+    --mlock --no-mmap \
+    --no-warmup \
+    --log-disable > /dev/null 2>&1