stardusts-hj

> @ygfrancois The position of [EOT] token is different for text with diff length, this dose not confuse the learning of position embedding? the [ETO] token is 49407 in this...

Another question is that since I only finetune the taesd decoder and leave the encoder freezed, should I also train the encoder of taesd with my datasets?

Thank you so much for your helpful suggestions and quick reply! I observe the degradation of images even during training so I might again check my training loss (1.a). For...

I also tried to build docker for vbench 2.0 and found the dependencies are too complicated, especially the mmcv. I launched several dockers, where all of them failed due to...