TensorRT_Deploy TRT's likely the fastest way to run a Deep learning model on NVIDIA GPUs. ^ . ^ Enviroments Ubuntu 20.04 GPU: RTX-3090 CPU: Intel Core i9-10940 X CUDA V11.1.105 cudnn 8.2.1 TensorRT 8.2.5.1 for RealESRGAN torch 1.8.1+cu111 Updates 2022/09/05 RealESRGAN_Quantization -> QAT and PTQ 2022/07/10 RealESRGAN_TRT -> TRT infer, support Int8 quantization PTQ by TRT. 2022/06/28 RealESRGAN_TRT -> TRT infer, support dynamic shape.