Assume there is a matrix H_W, and a matrix W_H, we will caculate matrix multiply, then how about the limitation with the H and W?
Link Copied
@thomascatlee The amount of information that can be processed on a NCS device is limited by the memory on the device. When considering maximum matrix size when using the NCS, there may be other factors can come into into play like the model's layers/operations being performed.
For more complete information about compiler optimizations, see our Optimization Notice.