Paolo Inglese

you need to use pytorch geometric 1.3.2

Please have a look at PR #71 I've optimized the speed and memory usage for `filterPeaks`. The bottleneck in that function is the binary matrix for peaks occurrence.

I'm having a look at the errors and fixing them.

Hi, I've made some modifications to the code. Since you make a distinction between 0 and NA, I've used a simple hack to drop the NAs and keep the 0....

> Wow, what a great PR! Thanks! :) > > Since you make a distinction between 0 and NA, I've used a simple hack to drop the NAs and keep...

> > In sparseMatrix, NAs occupy memory unfortunately. Otherwise, it would be straightforward to implement it. > > Thanks for clarification. I wasn't aware of this detail of the `Matrix`...

Hi @iiSeymour , here it is: ``` +-----------------------------------------------------------------------------------------+ | NVIDIA-SMI 550.54.14 Driver Version: 550.54.14 CUDA Version: 12.4 | |-----------------------------------------+------------------------+----------------------+ | GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC...

These are the last lines of the output before it crashes: ``` [2024-08-05 11:39:56.431] [info] cuda:0 using chunk size 12288, batch size 512 [2024-08-05 11:39:56.431] [debug] cuda:0 Model memory 36.54GB...

Just found out that one GPU is faulty, as the model works with the other 3 when used individually.

found the issue in one of my GPUs.