How to fit qwen3.8 27B on 5070 Ti 16GB VRAM

5070ti_qwen3.8_27B.sh

# gguf from https://huggingface.co/unsloth/Qwen3.8-27B-GGUF
# template jinja from https://huggingface.co/froggeric/Qwen-Fixed-Chat-Templates

./build/bin/llama-server \
    -m models/Qwen3.8-27B-UD-IQ4_XS.gguf \
    --alias "qwen3.8-27b-q4" \
    --mmproj models/mmproj-BF16.gguf \
    --no-mmproj-offload \
    --image-min-tokens 1024 \
    --image-max-tokens 1024 \
    --fit off \
    -ngl 999 \
    -c 64512 \
    -ctk q5_0 \
    -ctv q5_0 \
    -fa on \
    -np 1 \
    -b 256 \
    -ub 128 \
    --cache-reuse 64 \
    -t 12 \
    --jinja \
    --chat-template-file models/qwen-fixed-chat-template.jinja \
    --chat-template-kwargs '{"reasoning_effort":"medium"}' \
    --reasoning on \
    --reasoning-budget 8192 \
    --temp 1.0 \
    --top-p 0.95 \
    --top-k 20 \
    --min-p 0.0 \
    --presence-penalty 0.0 \
    --repeat-penalty 1.0 \
    --host 127.0.0.1 \
    --cors-origins localhost \
    --port 8080