Optimize OCR Model

[ming1753]

export FD_ENABLE_MAX_PREFILL=1
python -m fastdeploy.entrypoints.openai.api_server \
        --model /path/to/checkpoint \
        --port 8185 \
        --metrics-port 8189 \
        --engine-worker-queue-port 8198 \
        --cache-queue-port 55660 \
        --max-model-len 16384 \
        --max-num-batched-tokens 16384 \
        --gpu-memory-utilization 0.7 \
        --max-num-seqs 256 \
        --workers 2 \
        --graph-optimization-config '{"graph_opt_level":0, "use_cudagraph":true}' \

[paddle-bot]

Thanks for your contribution!