I’d go with a remote dev solution. Training/finetuning of large models requires much more resources anyway, so the GPUs in the local machine would be unused mos

I’d go with a remote dev solution. Training/finetuning of large models requires much more resources anyway, so the GPUs in the local machine would be unused most of the time.