README.md

metadata

license: apache-2.0

T5-XXL Encoder

This repo contains copies of the T5-XXL encoder in various quantization formats. The models in this repo are intended for use in InvokeAI.

Contents:

• bfloat16/ - T5-XXL encoder cast to bfloat16. Copied from here.
• bnb_llm_int8/ - T5-XXL encoder quantized using bitsandbytes LLM.int8() quantization.
• optimum_quanto_qfloat8/ - T5-XXL encoder quantized using optimum-quanto qfloat8 quantization.