2 ms·
I still don't think I'm getting my point across. What if you store 40% of the data and 40% of the metadata in CPU memory. Is that the same for performance? I
by Dylan16807 3y ago
I still don't think I'm getting my point across.
What if you store 40% of the data and 40% of the metadata in CPU memory. Is that the same for performance?
Is there a reason we want particular parts of the layer to stay on the GPU? Or is it just number of bytes.
- danielhanchen 3y agoOh, very good question - tbh im not sure. Another close technique is layer offloading - if your network can't fit and has layers 1, 2, ..., 32, we offload layers 16 to 32 to RAM, then load them in to GPU memory on the fly. I'm gonna guess the performance hit is similar - although I have not tried it myself to verify for benchmarking