OpenGVLab DiffRate issues

OpenGVLab / DiffRate

[ICCV 23]An approach to enhance the efficiency of Vision Transformer (ViT) by concurrently employing token pruning and token merging techniques, while incorporating a differentiable compression rate.

78 stars 7 forks source link

issues

Newest

Newest Most commented Recently updated Oldest Least commented Least recently updated

Inquiry on Fine-tuning Details for Table 2 in Your Repository

#5 SunkiLin opened 1 month ago
1
The prune/merge token numbers for the last layer is out of range.

#4 Bostoncake closed 1 month ago
1
Why is DiffRate only trained on MAE pre-trained ViT models?

#3 Bostoncake closed 1 month ago
1
Compression rate searching for a finetuned model

#2 Iambestfeed opened 7 months ago
4
How does the gradient flow into the FLOP loss?

#1 kaikai23 closed 1 month ago
1

OpenGVLab / DiffRate

issues

Inquiry on Fine-tuning Details for Table 2 in Your Repository

The prune/merge token numbers for the last layer is out of range.

Why is DiffRate only trained on MAE pre-trained ViT models?

Compression rate searching for a finetuned model

How does the gradient flow into the FLOP loss?