Closed rsdubtso closed 7 years ago
The original driver for MKL-DNN computed both gradients for weights and bias for BWD_F. This patch disables bias gradient computations by default. The aim of this change is to make it easier to compare MKL and MKL-DNN performance.
Looks OK to me, merged.
The original driver for MKL-DNN computed both gradients for weights and bias for BWD_F. This patch disables bias gradient computations by default. The aim of this change is to make it easier to compare MKL and MKL-DNN performance.