Added support for Tensor Cores CC >= 7.0 (V100). For FP16/32 (mixed precision) define CUDNN_HALF should be used.

This commit is contained in:
AlexeyAB
2018-02-25 16:29:44 +03:00
parent 85eafd3d59
commit cad4d1618f
8 changed files with 123 additions and 18 deletions

View File

@ -83,6 +83,8 @@ void free_layer(layer l)
if (l.x_norm_gpu) cuda_free(l.x_norm_gpu);
if (l.weights_gpu) cuda_free(l.weights_gpu);
if (l.weight_updates_gpu) cuda_free(l.weight_updates_gpu);
if (l.weights_gpu16) cuda_free(l.weights_gpu16);
if (l.weight_updates_gpu16) cuda_free(l.weight_updates_gpu16);
if (l.biases_gpu) cuda_free(l.biases_gpu);
if (l.bias_updates_gpu) cuda_free(l.bias_updates_gpu);
if (l.scales_gpu) cuda_free(l.scales_gpu);