'm experiencing extremely slow training times (multiple days) when running InstructLab with the following configuration:

Hardware: 12th Gen Intel Core i7, RTX 3060 (6GB VRAM), 64GB RAM
Software: InstructLab - 0.22.1

How did I execute it?

priscila@Hyrule ~ % ilab data generate --model ~/.cache/instructlab/models/merlinite-7b-lab-Q4_K_M.gguf --pipeline simple --gpus 1
priscila@Hyrule ~ % ilab model train --pipeline simple  --data-path /home/priscila/.local/share/instructlab/datasets/

It appears that InstructLab is utilizing only a single thread and minimal memory during the training process. I suspect this might be due to the limited VRAM available on my RTX 3060, despite having 64GB of system RAM and a powerful 12th Gen i7 CPU.

I have some questions:

• Is single-threaded execution the default behavior for InstructLab, or is there a configuration setting I'm missing to enable multi-threading?

• Are there any known workarounds or optimizations for running InstructLab with limited VRAM, especially when system RAM is abundant and a multi-core CPU is available?

• Is there a way to leverage the available system RAM and CPU cores to compensate for the limited VRAM?

Additional Information:

InstructLab version: 0.22.1
Operating System:  Fedora 41
Python version: 3.11