Hey, I'm getting deeper into model finetuning and training. I was just curious what most practitioners here prefer - do you invest in your own GPUs or rent compute when needed? Personally, I’ve grown frustrated with renting GPUs on platforms, but setting up my own environment keeps giving me errors. I wasted like 3 hours just fixing CUDA. I’m looking for a more integrated platform ,ideally with transparent pricing so I can control costs. Would love to hear what worked best for you and why.
[link] [comments]