search

dvlab-research / PFENet

PFENet: Prior Guided Feature Enrichment Network for Few-shot Segmentation (TPAMI).
322 stars 53 forks source link

about scale_lr #17

Closed roywithfiringblade closed 3 years ago

roywithfiringblade commented 3 years ago

In the experiment setting of PASCAL-5i, the initial lr is setting to 0.0025. But i saw that there was a scaler_lr in the poly_learning_rate(). And the default value is 10. So the initial lr is actually 0.025. right?

tianzhuotao commented 3 years ago

@704051419 In train.py, the index has been set as -1 which means the initial lr is scaled for all learnable parameters.

Details can be found at: https://github.com/Jia-Research-Lab/PFENet/blob/master/train.py#L255 https://github.com/Jia-Research-Lab/PFENet/blob/master/util/util.py#L49

Thank you for mentioning it and sorry for the inconvenience.