cntk.misc.optimized_rnnstack_converter module¶

convert_optimized_rnnstack(cudnn_model)[source]¶

Converts model that contains cudnn optimized_rnnstack to use non-cudnn functions, so it can be used in non-CUDA environment

Parameters:	cudnn_model – a model that contains optimized_rnnstacks
Returns:	Converted model on GEMM based implementation of rnn that can be used on CPU