Could you add some instructions on how to fine-tune the pretrained model in both step 1 and step 2?

[jocskywalker]

Hi,
Could you add some instructions on how to fine-tune the pretrained model in both step 1 and step 2? In step 1, I tried the config below: python main.py --model deit_tiny_patch16_224 --data-path /mnt//ViTCoD-main/data/CLS-LOC --lr 1e-5 --weight-decay 1e-8 --epochs 100 --resume exp_hr/svd/deit_tiny/deit_tiny_patch16_224-a1311bcf.pth --output_dir exp_hr/lowrank_sparse/deit_tiny/deit_tiny_info50 --svd_type 'mix_head_fc_qk' --need_weight , but get a very low accuracy in imagenet1k, the pretrained model is download from the DeiT official Lib. And should I use the gen_mask to create sparse mask before the step 2 training?

Thanks in advance