Hello,
I know this is maybe not the right place to ask this question. However, I think many people (with extensive experience) can be found in this forum, and they can help with some advice about what platform they use to train their Pytorch models in computer vision.
I wonder what platform you currently use to train machine learning models for computer vision.
I have tried Linode and GCP (Google Cloud Plattform) with good results. However, I find it rare that Linode (which provides an RTX Quatro 4000) offers the same speed (or even faster) than GCP (using its A100). This makes me think that other unspecified hardware (such as the brand of RAM or CPU) comes into the equation.
So, I started looking at other platforms, like Paperspace, Lamda, and Amazon, to train my models quickly and fast with the best cost. I am not interested in deployment services or even databases.
I am open to discussing your recommendations and personal opinion. I think this thread can be very productive and helpful for other people since the goal is to be able to develop our projects with Pytorch.
PS: I have categorized this topic under Vision because GPU RAM specs may differ from NPL models for vision models.