Are you sure you want to delete this task? Once this task is deleted, it cannot be recovered.
You can not select more than 25 topics
Topics must start with a chinese character,a letter or number, can include dashes ('-') and can be up to 35 characters long.
daquexian
9395a01939
|
4 years ago | |
|---|---|---|
| .. | ||
| all_v7 | 4 years ago | |
| arm64-v8a | 5 years ago | |
| arm64-v8a-wo-test | 5 years ago | |
| bconv_test | 5 years ago | |
| bconv_test_x86 | 5 years ago | |
| benchmark | 5 years ago | |
| benchmark_macro | 5 years ago | |
| benchmark_v7 | 4 years ago | |
| bgemm_test | 5 years ago | |
| bgemm_test_v7 | 4 years ago | |
| bgemm_test_x86 | 5 years ago | |
| bitpack_test | 5 years ago | |
| converter | 5 years ago | |
| converter_python | 5 years ago | |
| fconv_test | 5 years ago | |
| gemm_test | 5 years ago | |
| im2col_test | 5 years ago | |
| jni | 5 years ago | |
| net_test | 4 years ago | |
| net_test_debug | 5 years ago | |
| net_test_v7 | 4 years ago | |
| net_test_x86 | 5 years ago | |
| pad_test | 5 years ago | |
| run_net | 5 years ago | |
| run_net_debug | 5 years ago | |
| run_net_new_bitpack | 5 years ago | |
| run_net_x86 | 4 years ago | |
我们推出了 dabnn，一个用 ARM 汇编重写了卷积，高度优化的二值网络 inference 框架。实验显示 dabnn 相对于现有的二值网络和浮点网络 inference 框架均能带来极大的加速。在运行单个二值卷积时，我们的速度是 BMXNet （除 dabnn 外唯一一个二值网络 inference 框架） 的 800%~2400%，在运行二值化的 ResNet-18 时，我们的速度是 BMXNet 的约 700%。
C++ C Markdown Text Shell other