Ben出没冇注意的博客

问题描述：使用TensorFlow Model Optimization 0.6.0对自己的网络进行剪枝的时候，遇到官方暂时不支持的层ValueError: Please initialize with a supported layer. Layers should either be supported by the PruneRegistry (built-in keras layers) or should be a instance, or should has a customer def

Tensorflow2.9.1版本中加入的count flops的API

pytorch中的手册：torch.nn.KLDivLosskl_loss = nn.KLDivLoss(reduction="batchmean")# input should be a distribution in the log spaceinput = F.log_softmax(torch.randn(3, 5, requires_grad=True))# Sample a batch of distributions. Usually this would come from the

loss = y_true * log(y_true / y_pred)情况一：算两个一维数组参考：tf.keras.losses.KLDivergencek = tf.keras.losses.KLDivergence()loss = k([.4, .9, .2], [.5, .8, .12])print('Loss: ', loss.numpy()) # Loss: 0.11891246情况二：按batch来算参考：TensorFlow->API->TensorFlow

tf.signal.ffttf.signal.fft( input, name=None)根据最里面的维度计算一维离散傅里叶变换(DFT)参数Input : complex64和complex128 类型的张量(Tensor)name : 操作的名字(可选)返回与Input相同类型的一个Tensor.使用方法：a = tf.ones([4])a_complex = tf.complex(a,tf.zeros(a.shape)) #先构造一个复数b = tf.signal.

报错信息：Error polling for event status: failed to query event: CUDA_ERROR_ILLEGAL_INSTRUCTION: an illegal instruction was encountered2022-03-24 23:32:13.170887: F tensorflow/core/common_runtime/gpu/gpu_event_mgr.cc:273] Unexpected Event status: 1情况描述：我自定.

numpy.angle角度值的取值范围就是跟正弦函数sin一样

如图，金士顿DataTraveler 读文件时候发现经常出现错误，用DiskGenuis一看发现很多坏块，打算重新量产。先用ChipGenuis看一眼主控是群联PS2251-07的主控，固件版本是05.00.50 海力士TLC颗粒在http://www.upantool.com/liangchan/Phison/) 和upan.cc(http://www.upan.cc/tools/mass/Phison找了半天量产工具都不合适，报错。然后参考https://blog.youkuaiyun.com.