Score
0
30 Unstar Star 121 Fork 193

MindSpore / docsApache-2.0

Create your Gitee Account
Explore and code with more than 5 million developers,Free private repositories !:)
Sign up
  • Jupyter Notebook 95.1%
  • Python 4.7%
  • Makefile 0.1%
  • Shell 0.1%
Clone or download
benchmark.md 4.09 KB
Copy Edit Web IDE Raw Blame History
JunYuLiu authored 2020-09-11 11:36 . Modify the format of markdown

基准性能

Linux Ascend 模型训练 中级 高级

本文介绍MindSpore的基准性能。MindSpore网络定义可参考Model Zoo

训练性能

ResNet

Network Network Type Dataset MindSpore Version Resource                 Precision Batch Size Throughput Speedup
ResNet-50 v1.5 CNN ImageNet2012 0.5.0-beta Ascend: 1 * Ascend 910 CPU:24 Cores Mixed 256 2115 images/sec -
Ascend: 8 * Ascend 910 CPU:192 Cores Mixed 256 16600 images/sec 0.98
Ascend: 16 * Ascend 910 CPU:384 Cores Mixed 256 32768 images/sec 0.96
  1. 以上数据基于华为云AI开发平台ModelArts测试获得,是训练过程整体下沉至Ascend 910 AI处理器执行所得的平均性能。
  2. 业界其他开源框架数据可参考:ResNet-50 v1.5 for TensorFlow

BERT

Network Network Type Dataset MindSpore Version Resource                 Precision Batch Size Throughput Speedup
BERT-Large Attention zhwiki 0.5.0-beta Ascend: 1 * Ascend 910 CPU:24 Cores Mixed 96 269 sentences/sec -
Ascend: 8 * Ascend 910 CPU:192 Cores Mixed 96 2069 sentences/sec 0.96
  1. 以上数据基于华为云AI开发平台ModelArts测试获得,其中网络包含24个隐藏层,句长为128个token,字典表包含21128个token。
  2. 业界其他开源框架数据可参考:BERT For TensorFlow

Wide & Deep (数据并行)

Network Network Type Dataset MindSpore Version Resource                 Precision Batch Size Throughput Speedup
Wide & Deep Recommend Criteo 0.6.0-beta Ascend: 1 * Ascend 910 CPU:24 Cores Mixed 16000 796892 samples/sec -
Ascend: 8 * Ascend 910 CPU:192 Cores Mixed 16000*8 4872849 samples/sec 0.76
  1. 以上数据基于Atlas 800测试获得,且网络模型为数据并行。
  2. 业界其他开源框架数据可参考:Wide & Deep For TensorFlow

Wide & Deep (Host-Device混合计算模型并行)

Network Network Type Dataset MindSpore Version Resource                 Precision Batch Size Throughput Speedup
Wide & Deep Recommend Criteo 0.6.0-beta Ascend: 1 * Ascend 910 CPU:24 Cores Mixed 8000 68715 samples/sec -
Ascend: 8 * Ascend 910 CPU:192 Cores Mixed 8000*8 283830 samples/sec 0.51
Ascend: 16 * Ascend 910 CPU:384 Cores Mixed 8000*16 377848 samples/sec 0.34
Ascend: 32 * Ascend 910 CPU:768 Cores Mixed 8000*32 433423 samples/sec 0.20
  1. 以上数据基于Atlas 800测试获得,且网络模型为模型并行。
  2. 业界其他开源框架数据可参考:Wide & Deep For TensorFlow
1
https://gitee.com/mindspore/docs.git
git@gitee.com:mindspore/docs.git
mindspore
docs
docs
master

Search

131423 f1aaba0b 1899542 094922 1c74bed3 1899542