mirror of
https://github.com/pjreddie/darknet.git
synced 2023-08-10 21:13:14 +03:00
Fixed RNN (LSTM, RNN, CRNN, GRU) for CUDNN_HALF=1
This commit is contained in:
@ -215,6 +215,16 @@ void forward_crnn_layer_gpu(layer l, network_state state)
|
||||
layer input_layer = *(l.input_layer);
|
||||
layer self_layer = *(l.self_layer);
|
||||
layer output_layer = *(l.output_layer);
|
||||
/*
|
||||
#ifdef CUDNN_HALF
|
||||
// slow and bad
|
||||
s.index = state.index;
|
||||
s.net = state.net;
|
||||
cuda_convert_f32_to_f16(input_layer.weights_gpu, input_layer.c*input_layer.n*input_layer.size*input_layer.size, input_layer.weights_gpu16);
|
||||
cuda_convert_f32_to_f16(self_layer.weights_gpu, self_layer.c*self_layer.n*self_layer.size*self_layer.size, self_layer.weights_gpu16);
|
||||
cuda_convert_f32_to_f16(output_layer.weights_gpu, output_layer.c*output_layer.n*output_layer.size*output_layer.size, output_layer.weights_gpu16);
|
||||
#endif //CUDNN_HALF
|
||||
*/
|
||||
|
||||
fill_ongpu(l.outputs * l.batch * l.steps, 0, output_layer.delta_gpu, 1);
|
||||
fill_ongpu(l.hidden * l.batch * l.steps, 0, self_layer.delta_gpu, 1);
|
||||
|
Reference in New Issue
Block a user