3 ms·
Yes, but the GPU is needed for training so "we" do this on prem or in the cloud. But the actual prediction computations are fast. Do we really need an edge GP
by jgalt212 7y ago
Yes, but the GPU is needed for training so "we" do this on prem or in the cloud. But the actual prediction computations are fast. Do we really need an edge GPU for these?
- TaylorAlexander 7y agoYes. An edge GPU is required to run inference on high resolution sensor data for many robotics applications. Running inference is faster than training but it’s not all “fast”. Inference on CPU can take many seconds with large image operations. Not useful for certain robotics applications.