mirror of https://github.com/AlexeyAB/darknet.git
Bug fix
This commit is contained in:
parent
90c15f8f04
commit
9c9232d1c3
|
@ -1429,7 +1429,6 @@ network parse_network_cfg_custom(char *filename, int batch, int time_steps)
|
||||||
while(n){
|
while(n){
|
||||||
|
|
||||||
if (count < last_stop_backward) params.train = 0;
|
if (count < last_stop_backward) params.train = 0;
|
||||||
else params.train = 1;
|
|
||||||
|
|
||||||
params.index = count;
|
params.index = count;
|
||||||
fprintf(stderr, "%4d ", count);
|
fprintf(stderr, "%4d ", count);
|
||||||
|
@ -1693,9 +1692,9 @@ network parse_network_cfg_custom(char *filename, int batch, int time_steps)
|
||||||
for (k = 0; k < last_stop_backward; ++k) {
|
for (k = 0; k < last_stop_backward; ++k) {
|
||||||
layer l = net.layers[k];
|
layer l = net.layers[k];
|
||||||
if (l.keep_delta_gpu) {
|
if (l.keep_delta_gpu) {
|
||||||
if (!l.delta) l.delta = (float*)xcalloc(l.outputs, sizeof(float));
|
if (!l.delta) l.delta = (float*)xcalloc(l.outputs*l.batch, sizeof(float));
|
||||||
#ifdef GPU
|
#ifdef GPU
|
||||||
if (!l.delta_gpu) l.delta_gpu = (float *)cuda_make_array(NULL, l.outputs);
|
if (!l.delta_gpu) l.delta_gpu = (float *)cuda_make_array(NULL, l.outputs*l.batch);
|
||||||
#endif
|
#endif
|
||||||
}
|
}
|
||||||
|
|
||||||
|
|
Loading…
Reference in New Issue