Out of memory during training

[almosti]

I have a specific inquiry regarding the hardware specifications used during the training of the masif-site and masif-search models.

Despite utilizing a computer with 16GB of physical memory and an additional 16GB of virtual memory, I am consistently encountering issues related to insufficient memory during the training of MaSIF-site. In light of this, I am eager to gain insights into the CPU, GPU, and memory configurations that you employed for training these models. Additionally, I chose CPU-based training due to compatibility issues with the TensorFlow version. I am curious whether this decision might be contributing to the observed memory constraints and if you have any recommendations or insights to address this issue.

Any comment or ideas would be appreciated!