fanlamda

Yes，I used self.model:float() After cudnn.convert(), but I found cudnn.batchBRNNReLU remains unchanged. It comes error like 'unknown object cudnn.batchBRNNReLU'

I can understand why GPU menmory usage increase during a batch. But why does GPU memory not get back to as low as before. @SeanNaren

@riebling thanks for reply actually when I enter ./configure --shared --cudatk-dir=/path/to/cuda_library I got the following. It seems OK with CUDA Validating presence of ATLAS libs in /usr/lib Using library /usr/lib/liblapack.so...