Lots of logspam volume that I'm immediately going to tamp down.
But it's also got me thinking I'm going to buy a bunch of GPUs this month and stand up our own ML compute cluster.
We'll still run CRUD in the cloud, but I'm getting sick of paying for cloud GPUs. The GKE / GPU interface leaves tons of spare compute on the table anyway.
But it's also got me thinking I'm going to buy a bunch of GPUs this month and stand up our own ML compute cluster.
We'll still run CRUD in the cloud, but I'm getting sick of paying for cloud GPUs. The GKE / GPU interface leaves tons of spare compute on the table anyway.