'Retraining a Model from 3 Channels (RGB) to 4 Channels (RGBA), can I use the 3 channel weights?
I need to expand a model from RGB to RGBA. I can handle the code rewrite on the model, but instead of retraining the entire model from scratch, I would love to start it off with it's 3 channel weights + zeros.
Is there an easy way to change torch's save of 3 channel weights into 4?
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|