Egill Axfjord Fridgeirsson

Hi @dfalbel , > The main reason for not using the C++ implementations was that I wanted to allow extensions in the R side, and It would tricky to call...

Hi @dfalbel, That was quick! But I can't open the repo, is it by any chance private?

@dfalbel some preliminary results for a small ResNet run for 20 epochs on the same random data With torch R Adam optimizer: ```Average time per epoch was: 0.854 secs``` With...

> That shouldn't be a problem though because in our implementation LibTorch will call R's gc if it's running out of GPU memory: I've actually been sometimes running out of...

Unfortunately the solution of adding ```gc(full = TRUE)``` to the training loop causes quite a slowdown, more than 2x. Using the full-gc branch (not sure it was ready for testing...

It's definitely better in terms of memory, and the performance hit is not as severe, it's mostly now during the first epoch. The memory use is still higher than in...

Thanks for helping me understand all this. I was looking at nvidia-smi, which you already told me wasn´t the best but I guess it's an old habit. Would you suggest...

Hi @iamalonso, sorry for the late response. What kind of features do you have? I made a short snippet here where I develop a model and then use it to...

What about having it as a dataframe? Seems more appropriate for this kind of 2d data. That's what I did to test, stored it as dataframe before it was written...

Hi @ablack3 , @azimov , @ginberg I have made a reprex in the following docker container: https://hub.docker.com/r/egillax/shiny_debug_java `docker run -ti -p 3838:3838 egillax/shiny_debug_java` See app details at bottom. I had...