issues
search
jeongjae96
/
AI
인공지능 관련 학습을 위한 repository
0
stars
0
forks
source link
[docs] add explanation for placing the stride at 3x3 convolution for bottleneck #8
#10
Closed
jeongjae96
closed
1 year ago
jeongjae96
commented
1 year ago
improves accuracy according to
https://ngc.nvidia.com/catalog/model-scripts/nvidia:resnet_50_v1_5_for_pytorch