binary): 1.5.0 -- Python version (e.g., Python 3.7.5): Python 3.7.5 -- OS platform and distribution (e.g., Linux Ubuntu 16.04): Ubuntu

"avatar": "https://cdn.jsdelivr.net/gh/kelecn/images@master/%E6%82%9F%E7%A9%BA.jpg", "screenshot": "https://cdn.jsdelivr.net/gh/k

349 757 106 355] imag_count 50000, top1_accuracy 0.000 top5_accuracy 0.003 二、软件版本: -- CANN 版本 (e.g., CANN 3.0.x，5.x.x): CANN 5.0.2

(binary): 1.2.0 -- Python version (e.g., Python 3.7.5): -- OS platform and distribution (e.g., Linux Ubuntu 18.04): -- GCC/Compiler

cross_entropy(input=out, label=label) acc_top1 = fluid.layers.accuracy(input=out, label=label, k=1) acc_top5 = fluid.layers.accuracy(input=out

https://tinysu.gitee.io/2020/06/15/%E5%8A%9B%E6%89%A3%E7%B2%BE%E9%80%89TOP%E9%9D%A2%E8%AF%95%E9%A2%98(41-50)/

用于计算db_multiplier。（默认值：1.0） | | amin | float | 用于对输入的音频进行最小值的截断。（默认值：1e-10） | | top_db | float | 最小负截至分贝。通常设为80。（默认值：80） |

nums; } private int top; public int getTop() { return top; } public void setTop(int top) { this.top = top; } public Stack(){};

4、sdk推理精度低于模型训练精度0.55%，推理精度不达标： sdk推理精度：Top1:58.45% Top5:81.34% B类交付模型评估精度：Top1:59% Top5:81.84% 5、modelarts训练需用户修改代码才能适配迁移学习训练，建议适配代码避免让用户来修改代码：

https://tinysu.gitee.io/2020/06/15/%E5%8A%9B%E6%89%A3%E7%B2%BE%E9%80%89TOP%E9%9D%A2%E8%AF%95%E9%A2%98(21-30)/

https://tinysu.gitee.io/2020/06/15/%E5%8A%9B%E6%89%A3%E7%B2%BE%E9%80%89TOP%E9%9D%A2%E8%AF%95%E9%A2%98(11-20)/