diff --git a/cfg/darknet.cfg b/cfg/darknet.cfg index 85497a37..00e9c366 100644 --- a/cfg/darknet.cfg +++ b/cfg/darknet.cfg @@ -7,10 +7,11 @@ channels=3 momentum=0.9 decay=0.0005 -learning_rate=0.1 -policy=poly -power=4 -max_batches=500000 +learning_rate=0.01 +policy=sigmoid +gamma=.00002 +step=400000 +max_batches=800000 [crop] crop_height=224 diff --git a/src/convolutional_kernels.cu b/src/convolutional_kernels.cu index 130824af..a64a499e 100644 --- a/src/convolutional_kernels.cu +++ b/src/convolutional_kernels.cu @@ -122,6 +122,12 @@ void forward_convolutional_layer_gpu(convolutional_layer l, network_state state) fast_mean_gpu(l.output_gpu, l.batch, l.n, l.out_h*l.out_w, l.mean_gpu); fast_variance_gpu(l.output_gpu, l.mean_gpu, l.batch, l.n, l.out_h*l.out_w, l.variance_gpu); + /* + cuda_pull_array(l.variance_gpu, l.mean, 1); + printf("%f\n", l.mean[0]); + */ + + scal_ongpu(l.n, .95, l.rolling_mean_gpu, 1); axpy_ongpu(l.n, .05, l.mean_gpu, 1, l.rolling_mean_gpu, 1); scal_ongpu(l.n, .95, l.rolling_variance_gpu, 1);