# Since `swift/test_lora` is trained by swift and contains an `args.json` file,
# there is no need to explicitly set `--model`, `--system`, etc., as they will be automatically read.
CUDA_VISIBLE_DEVICES=0 swift deploy \
    --host 0.0.0.0 \
    --port 8000 \
    --adapters swift/test_bert \
    --served_model_name bert-base-chinese \
    --infer_backend pt \
    --truncation_strategy right \
    --max_length 512