SDXL CLIP doesn't support token length over 77

[Anmuliar]

Since sometimes the regional prompt generated by RPG will be over the capacity of the sdxl CLIP (can only handle sequence up to 77 tokens), is there any alternative model can cope with such a situation. I used RPG.py with GPT-4 and sdxl-1.0-base. Or have I missed some settings in the RPG code.