Indeed EnsembleGPUArray is the wrong thing to use here. In theory you can get this to work but it won’t be as fast as batching the normal way, so I’d just GPU accelerate the neural network calls the normal way here shown in the docs.
Indeed EnsembleGPUArray is the wrong thing to use here. In theory you can get this to work but it won’t be as fast as batching the normal way, so I’d just GPU accelerate the neural network calls the normal way here shown in the docs.