start.sh

nohup python3 gradio_demo/seed_llama_gradio.py --server_port 7860 --request_address http://127.0.0.1:7890/generate --model_type seed-llama-14b >./output.out &
python3 gradio_demo/seed_llama_flask.py \
    --image_transform configs/transform/clip_transform.yaml \
    --tokenizer configs/tokenizer/seed_llama_tokenizer_hf.yaml \
    --model configs/llm/seed_llama_14b_8bit.yaml \
    --port 7890 \
    --llm_device cuda:0 \
    --tokenizer_device cuda:0 \
    --offload_decoder