Or is single precision (FP32) sufficient for CPUs & GPUs?
I’m using N-series Azure instances & there’s a single precision option which is cheaper.
You already have an issue open with this topic. I’ll reply there.
Or is single precision (FP32) sufficient for CPUs & GPUs?
I’m using N-series Azure instances & there’s a single precision option which is cheaper.
You already have an issue open with this topic. I’ll reply there.