Don’t BLINDLY reuse pre-trained models from a hub.
Nov 21, 2022
→ Why? The models can be,
- Over-trained
- Under-trained
- Over-parameterized
All of this can negatively impact your transfer or compression process like Teacher-student KD before transfer.
→ Frameworks supported
- Tensorflow 2.x / Keras
- PyTorch 1.x
→ Use WeightWatchers (WW)
An open-source, diagnostic tool for analyzing Deep Neural Networks (DNN), without needing access to training or even test data.
→ Easy to use, Snippet below
Repo: https://lnkd.in/eCg9jKHT