Emirhan Kurtuluş
Emirhan Kurtuluş
When I was working with Knet, I felt the need of classification metrics for a better evaluation of my model, but saw that Knet does not offer such a functionality,...
### System Info I am trying to replicate T5 finetuning on WMT with the following hyperparameters (as close as possible to the paper https://www.jmlr.org/papers/volume21/20-074/20-074.pdf): --model_name_or_path t5-small --source_lang en --target_lang de...
With this update, I am aiming for generalizing our training framework. I need feedback and review on the structure that I propose. Accordingly, I will make the codebase fully functional...
Thank you very much for the great work. It is stated in the paper that for the linear evaluation protocol on finetuning datasets, LBFGS optimizer was used. For how many...
This PR implements [Tied-Augment](https://proceedings.mlr.press/v202/kurtulus23a/kurtulus23a.pdf) for #1828.
Recently, we introduced [Tied-Augment](https://arxiv.org/abs/2305.13520), a simple framework that combines self-supervised learning learning and supervised learning by making forward passes on two augmented views of the data with tied (shared) weights....
Why are there no labels in the provided .jsonl file ?
Are there any plans for Image2Image pipeline for the StableCascade model ?
Great work! When I was taking a look at your code and your example, I saw no mention of mixed precision. Does the current implementation of SAM and GSAM support...
Please take a look at the lines 157-158 and 202