models:
  "DeepSeek-R1-0528-Qwen3-8B-GGUF":
    cmd: |
      /app/llama-server
      -hf unsloth/DeepSeek-R1-0528-Qwen3-8B-GGUF:Q4_K_M
      -ngl 37 -c 16384
      --port ${PORT}
  "Qwen3-8B-GGUF":
    cmd: |
      /app/llama-server
      -hf unsloth/Qwen3-8B-GGUF:Q4_K_M
      -ngl 37 -c 16384
      --port ${PORT}
  "Qwen3-8B-GGUF-no-thinking":
    cmd: |
      /app/llama-server
      -hf unsloth/Qwen3-8B-GGUF:Q4_K_M
      -ngl 37 -c 16384
      --jinja --chat-template-file /config/qwen_nothink_chat_template.jinja
      --port ${PORT}
  "gemma3n":
    cmd: |
      /app/llama-server
      -hf unsloth/gemma-3n-E4B-it-GGUF:UD-Q4_K_XL
      --ctx-size 16384
      --n-gpu-layers 99
      --seed 3407
      --prio 2
      --temp 1.0
      --repeat-penalty 1.0
      --min-p 0.00
      --top-k 64
      --top-p 0.95
      --port ${PORT}