Maximilian Pichler

Profiles from r-torch and imported torch:   Beside the fact that all steps are slower than the imported python-torch, the step method from the optimizer is disproportionaley slower than...

Hi @dfalbel, thanks for your quick response! Yeah, I played around and for larger problems there is already no notable difference between r-torch and imported python torch.

update 02/2021: I repeated the benchmarks with the current development version (altough my hardware has changed, but since we are more interested in relative runtimes between the different options that...

update 04/2021: (my hardware has changed again...)  In short, r-torch/imported-torch: 4.924 -> 3.661 -> 3.369 Also, the advantage (overhead) of native pytorch over r-torch decreased von 19x to 12x...

Hi, The model tried to allocate 60GB, but your GPU has only 48GB. Possible solutions to reduce the memory consumption of your model - Decrease the step_size - Decrease sampling...

`step_size` is automatically set to 10% of your data (which can consume a lot of memory for large data), so you might want to try it for your full data:...

Hi @Loic-sanchez , the anova can be very slow, the default for the MC samples in the anova is 5000 (samples = 5000L), maybe rerun it with only 100-500 samples:...

Hi @Loic-sanchez, Run `reticulate::use_condaenv("r-sjsdm", required = TRUE)` at the top of your script or in a new R session before loading a library.

hi @fetafretka, Not yet, we are still working on it. What kind of random factors would you like to include?

Okay, so how many sites do you have? If you have enough sites, you could run ~year*(env1+env2+...) + 0, which will give you independent slopes for env. Now, if you...