Created by: kuke

We are launching a cooperation with other team in speech recognition, and fluid will be chosen as the framework preferentially. To train models on large-scale audio data, we wonder when the multi-GPU support of fluid in one singe node will be ready.