Petru-Daniel Tudosiu

Results 11 issues of Petru-Daniel Tudosiu

Implementing FixUp initialization [1] based convolutional layer as it allows one to train without normalization. [1] Zhang, Hongyi, Yann N. Dauphin, and Tengyu Ma. "Fixup initialization: Residual learning without normalization."...

This issue was sparked by the implementation of SubPixel Convolution that is available in Fast.ai [1] Things to do: - [x] Restructure SubPixel Convolution since it does not fall in...

Is your feature request related to a problem? Please describe. When checkpoint saving occurs there should be a check if the object that state_dict() is called on is a ZeroRedundancyOptimizer...

Hi Lucidrains, I have stumbled across this paper that might be of interest to you. Given how transformers are memory/compute hungry I think this will be highly useful for the...

What are your thoughts about extending the library to include the following metrics: - 3-SSIM - 3-MS-SSIM - G-SSIM - G-MS-SSIM - 3-G-SSIM - 3-G-MS-SSIM - 4-SSIM - 4-MS-SSIM -...

There are two interesting features (low implementation overhead) from the paper [Robust Training of Vector Quantized Bottleneck Models](https://arxiv.org/pdf/2005.08520.pdf): - 3.A Importance of proper scaling - Batch normalisation - This is...

Hi Austin, I was wondering if it is possible to average across multiple runs/folds of the same model. I am trying at the moment, but the resulting ICE plots do...

**Is your feature request related to a problem? Please describe.** Create an inferer for aleatoric and epistemic uncertainty. **Describe the solution you'd like** The inferer can be based on the...

**Is your feature request related to a problem? Please describe.** Due to the complexities required by the creation of interfaces for the backend and frontend to interact the "information" flow...

backlog

**Is your feature request related to a problem? Please describe.** At the moment there is no straightforward way to decouple the Image Signal and Guidance Signal from the DeepEdit (and...

backlog