Hi, thanks for the amazing work you've done!
I follow the README file and reproduce the results of semantic segmentation LP and 1% fine-tuning on Nuscenes.
However, I got a much lower mIoU than proposed in your paper like LP: 8.52 and 1% fine-tuning: 22.9, are there any tricks in training?
The experiment was implemented on two 4090 gpus.
Hi, thanks for the amazing work you've done!
I follow the README file and reproduce the results of semantic segmentation LP and 1% fine-tuning on Nuscenes.
However, I got a much lower mIoU than proposed in your paper like LP: 8.52 and 1% fine-tuning: 22.9, are there any tricks in training?
The experiment was implemented on two 4090 gpus.