search

musco-ai / musco-pytorch

MUSCO: MUlti-Stage COmpression of neural networks
BSD 3-Clause "New" or "Revised" License
71 stars 16 forks source link

How to restore the ideal accuracy(mAP) by fine-tuning #6

Open Lawrencechengsjtu opened 4 years ago

Lawrencechengsjtu commented 4 years ago

Hi author! Thx for ur sharing!

I was just trying your iterative compression algorithms using vbmf for compressing the faster rcnn model (exactly the same code mentioned in your paper), but i found it great difficulty doing the fine-tuning work. The more layers I compressed, the less mAP it achieved. Finally, it is approximately 8~10 points lost, which is far below your performance.

Can u tell me how I should do the fine-tuning part better? (like dataset, lr, epoch, etc.) Or can u tell me some of your opinions in terms of it? Thank u!

Lawrencechengsjtu commented 4 years ago

@juliagusak

cszer commented 4 years ago
  1. compress backbone predtrained on imagenet firstly after fine tune on imagenet , it's most computational part of every detector. I think it's more appropriate approach