提交 ee0d365a 编写于 作者: T tensor-tang

add inference benchmark data

上级 8ed8a935
......@@ -23,6 +23,8 @@ On each machine, we will test and compare the performance of training on single
## Benchmark Model
### Server
#### Training
Test on batch size 64, 128, 256 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
Input image size - 3 * 224 * 224, Time: images/second
......@@ -62,6 +64,34 @@ TBD
chart on batch size 128
TBD
#### Inference
Test on batch size 1, 2, 4, 8, 16 on Intel(R) Xeon(R) Gold 6148 CPU @ 2.40GHz
- VGG-19
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|-------|-------|-------|-------|-------|
| OpenBLAS | 0.36 | 0.48 | 0.56 | 0.50 | 0.43 |
| MKLML | 5.41 | 9.52 | 14.71 | 20.46 | 29.35 |
| MKL-DNN | 65.52 | 89.94 | 83.92 | 94.77 | 95.78 |
- ResNet-50
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|-------|--------|--------|--------|--------|
| OpenBLAS | 0.29 | 0.43 | 0.71 | 0.85 | 0.71 |
| MKLML | 6.26 | 11.88 | 21.37 | 39.67 | 59.01 |
| MKL-DNN | 90.27 | 134.03 | 136.03 | 153.66 | 211.22 |
- GoogLeNet
| BatchSize | 1 | 2 | 4 | 8 | 16 |
|-----------|--------|--------|--------|--------|--------|
| OpenBLAS | 12.47 | 12.36 | 12.25 | 12.13 | 12.08 |
| MKLML | 22.50 | 43.90 | 81.22 | 132.92 | 199.69 |
| MKL-DNN | 221.69 | 341.33 | 428.09 | 528.24 | 624.18 |
### Laptop
TBD
### Desktop
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册