_taurus/ Hi3516dv300设备） OS版本：3.0LTS 用Image组件加载200k gif动图，使用 hdc_std shell top发现应用占了100%cpu，并且动图会出现如下残影，（不加载动图，应用占cpu 30%左右） ![

7/site-packages/paddle/v2/fluid/layers/nn.py", line 1032, in accuracy attrs={"k": k}) File "/home/ssd1/yancanxiang/code_paddle/se_resnext50/pyth

%0D%0Ainput%20must%20have%20%3E%3D%20k%20columns%20at%20%5B/paddle/paddle/fluid/operators/top_k_op.cc%3A40%5D%0D%0APaddlePaddle%20Call%20Stack

/www/wwwroot/store1.jerrywang.top/store find store -type d -exec mkdir -p /www/wwwroot/cdn.jerrywang.top/storage/ { } ; ``` ## 查找合适的deb并复制

reverse_dictionary[valid_examples[i]] top_k = 8 # number of nearest neighbors nearest = (-sim[i, :]).argsort()[1:top_k + 1] log_str = 'Nearest

com/PaddlePaddle/Paddle/issues/5202): An accuracy layer shall contain a `top_k` op and an `accuracy` op. It is needed to evaluate our training

sequence in the nested sequence. - Then sub_nest_seq_layer selects top k sentences with highest scores or probabilities according to the

e/issues/2585): 现象，集群训练报错，Local模式正常。期间使用classification_cost报top k，依照#2574方法尝试解决，报相关错误如下： Fri Jun 23 15:40:01 2017[1,53]:F0623 15:40:01

"generate.py", line 254, in main temperature=temperature, top_k=topk, top_p=topp, repitition_penalty=repetition_penalty, device=device

org/documentation/docs/zh/1.4/api_cn/layers_cn.html#topk 测试代码如下 ``` def test_top(): input = fluid.layers.data(name='input', shape=[5], dtype='float32')

def classification_cost(input, label, weight=None, name=None, top_k=None, evaluator=classification_error_evaluator, layer_attr=None):