luoganttcc的博客

import tensorflow as tfx = tf.constant([1, 2, 3])y = tf.broadcast_to(x, [5, 3])print(y)tf.Tensor([[1 2 3] [1 2 3] [1 2 3] [1 2 3] [1 2 3]], shape=(5, 3), dtype=int32)

计算给定符号形状的广播的形状,1.广播的原则如果两个数组的后缘维度（trailing dimension，即从末尾开始算起的维度）的轴长度相符，2.或其中的一方的长度为1，则认为它们是广播兼容的。广播会在缺失和（或）长度为1的维度上进行。张量的广播机制import tensorflow as tfshape_x = (6, 3)shape_y = (5, 1, 3)c=tf.broadcast_dynamic_shape(shape_x, shape_y)print(c)

tensorflow 的字面翻译是----张量(tensor)—流(flow)为什么要引入张量的广播机制，原因是有利可图，这个利益就是能节约内存空间，说白了就是省钱！！！凡事都有两面，既然能省钱，那一定费脑子，哲学一点的说法就是’抽象’对于张量的加法，乘法，不要求两个张量的维度完全相同，对于有很多重复元素的张量而言，可以用一个低纬度张量代表高维张量广播的原则如果两个数组的后缘维度（trailing dimension，即从末尾开始算起的维度）的轴长度相符，或其中的一方的长度为1，则认为它们是广播

对于tf.argmax,这个函数有点奇怪，axis=0指的是计算矩阵每列的最大值索引，axis=1计算行最大值索引与numpy 相同import tensorflow as tfimport numpy as npa=np.array([[2,4,5,7],[9,3,6,2]])print('-'*30+'分割线'+'-'*30)print(a)print('-'*30+'分割线'+'-'*30)a1=tf.argmax(a,axis=0)print('tf.argmax(a,ax

对于tf.reduce_max,这个函数有点奇怪，axis=0指的是计算矩阵每列的最大值，axis=1计算行最大值与numpy 不同import tensorflow as tfimport numpy as npa=np.array([[2,4,5,7],[9,3,6,2]])print('a=\n',a)print('-'*30+'分割线'+'-'*30)a1=tf.reduce_max(a,axis=0)print( 'tf.reduce_max(a,axis=0)=\n'

从张量的形状中删除大小为 1 的维度import tensorflow as tfa=tf.constant([[[1,2,3,1]]])print(a.shape)a1=tf.squeeze(a)print(a1.shape)(1, 1, 4)(4,)文档链接

Am×n×..×km \times n \times ..\times km×n×..×k 和A1×k1\times k1×k 矩阵相加相当于在最后一个维度import numpy as np a=np.array(range(3*4)).reshape([3,4])b=np.array([0.2]*4)print('a.shape=',a.shape)print('b.shape=',b.shape)print(a+b)print('-'*20+'我是分割线'+'-'*20)

githubimport tensorflow as tffrom absl.flags import FLAGS@tf.functiondef transform_targets_for_output(y_true, grid_size, anchor_idxs): #这个函数分别对比某一类anchors (一共是三类，每一类对应不同的尺寸的box) #每一类box 对应的尺寸翻倍 # y_true: (N, boxes, (x1, y1, x2, y2, cl

对应位置的索引赋值这里是一维坐标赋值import tensorflow as tftensor = [0, 0, 0, 0, 0, 0, 0, 0] # tf.rank(tensor) == 1indices =[ [1], [3],[4], [7]] # num_updates == 4, index_depth == 1updates = [ 9, 10, 11, 12] # num_updates == 4print(tf.

tensorflow 动态数组随时可以读取import tensorflow as tfta = tf.TensorArray(tf.float32, size=0, dynamic_size=True, clear_after_read=False)ta = ta.write(0, 10)ta = ta.write(1, 20)ta = ta.write(2, 30)print(ta.read(0))print(ta.read(1))print(ta.read(2))prin

import tensorflow as tftf.minimum([5,2,3],[2,3,4]) <tf.Tensor: shape=(3,), dtype=int32, numpy=array([2, 2, 3], dtype=int32)>tf.minimum([[3,6,7],[5,2,3]],[2,3,4])<tf.Tensor: shape=(2, 3), dtype=int32, numpy=array([[2, 3, 4], [2, 2,

import tensorflow as tfarr=tf.constant([[1,2,3],[4,5,6]])print(arr)print('-'*30)for k in range(3): cc1=tf.expand_dims(arr, axis=k) print(cc1) # print(cc1.shape.as_list()) print('*'*30)

tf.lookup.StaticHashTable 本质是tensorflow 内置字典，在yolov3 tf代码中多次应用def load_tfrecord_dataset(file_pattern, class_file, size=416): LINE_NUMBER = -1 # TODO: use tf.lookup.TextFileIndex.LINE_NUMBER class_table = tf.lookup.StaticHashTable(tf.lookup.TextF

tf.data.DatasetAPI支持写入的描述性和高效的输入管线。Dataset用法遵循一个常见模式：从输入数据创建源数据集。应用数据集转换来预处理数据。迭代数据集并处理元素。迭代以流式方式发生，因此不需要将完整数据集放入内存中。import tensorflow as tfdataset = tf.data.Dataset.from_tensor_slices([1, 2, 3])for element in dataset: print(element)dataset =

添加链接描述https://github.com/wziji/deep_ctr/tree/master/ESMM

import tensorflow as tfimport numpy as npelems = np.array([1, 2, 3, 4, 5, 6])sum = tf.scan(lambda a, x: a + x, elems)# sum == [1, 3, 6, 10, 15, 21]

t =[0, 1, 2, 3, 4]tf.roll(t, shift=2, axis=0) <tf.Tensor: shape=(5,), dtype=int32, numpy=array([3, 4, 0, 1, 2], dtype=int32)>这里的 axis=[0]比较奇怪，0轴表示列，1轴表示行t=[[0, 1, 2, 3, 4], [5, 6, 7, 8, 9]]t1=tf.roll(t, shift=[1], axis=[0])print('t1=',t1)t2

tf.reverse_sequence 只翻转前n个数据，seq_lengths = [7, 2, 3, 5]就是inptuts 第一行的前7个，第二行的前2个…import tensorflow as tfseq_lengths = [7, 2, 3, 5]inputs = [[1, 2, 3, 4, 5, 0, 0, 0], [1, 2, 0, 0, 0, 0, 0, 0], [1, 2, 3, 4, 0, 0, 0, 0], [1, 2, 3, 4, 5, 6, 7, 8]]

在之前的博客中我说过，一维度的矩阵是队列二维度的矩阵是方阵三维度的矩阵是大楼四维度的矩阵是小区我以三维矩阵来说 tf.reverse的意义，可以把它想象成一座大楼import tensorflow as tf t=tf.constant( [ [[ 0, 1, 2, 3], [ 4, 5, 6, 7], [ 8, 9, 10, 11]], [[ 10,

import tensorflow as tft=tf.constant([1, 2, 3, 4, 5, 6, 7, 8, 9])tf.reshape(t, [3, 3]) t=tf.constant( [ [[1, 1], [2, 2]], [[3, 3], [4, 4]] ])# tensor 't' has shape [2, 2, 2]tf.reshape(t, [2, 4]) <tf.Tensor: shape=(2, 4), dtype=int

import tensorflow as tftf.repeat(['a', 'b', 'c'], repeats=[3, 0, 2], axis=0)<tf.Tensor: shape=(5,), dtype=string, numpy=array([b'a', b'a', b'a', b'c', b'c'], dtype=object)>repeat([[1, 2], [3, 4]], repeats=[2, 3], axis=0)<tf.Tensor: shape=(5

import tensorflow as tf# shape of tensor 't' is [2, 2, 3]t = tf.constant([[[1, 1, 1], [2, 2, 2]], [[3, 3, 3], [4, 4, 4]]])tf.rank(t) # 3张量的秩与矩阵的秩不一样.张量的秩是唯一选择张量的每个元素所需的索引的数量.可以理解为维度秩也被称为 “order”,“degree” 或 “ndims”....

tf.ones([3, 4], tf.int32)<tf.Tensor: shape=(3, 4), dtype=int32, numpy=array([[1, 1, 1, 1], [1, 1, 1, 1], [1, 1, 1, 1]], dtype=int32)>def make_variables(k, initializer): return (tf.Variable(initializer(shape=[k], dtype=tf.float32)

通过继承tf.Module代替object任何tf.Variable或tf.Module分配给对象的属性的实例可以使用被收集variables ， trainable_variables或submodules属性：import tensorflow as tf class Dense(tf.Module): def __init__(self, in_features, out_features, name=None): super(Dense, self).__init__..

import numpy as npimport tensorflow as tfelems = np.array([1, 2, 3, 4, 5, 6])tf.map_fn(lambda x: x * x, elems)<tf.Tensor: shape=(6,), dtype=int64, numpy=array([ 1, 4, 9, 16, 25, 36])>

import tensorflow as tfelems = tf.constant([1, 2, 3, 4, 5, 6])sum1 = tf.foldl(lambda a, x: a * x, elems)print(sum1.numpy())sum2 = tf.foldl(lambda a, x: a + x, elems)print(sum2.numpy())

tf.fill([2, 3], 9)<tf.Tensor: shape=(2, 3), dtype=int32, numpy=array([[9, 9, 9], [9, 9, 9]], dtype=int32)>

tf.eye(2)<tf.Tensor: shape=(2, 2), dtype=float32, numpy=array([[1., 0.], [0., 1.]], dtype=float32)>tf.eye(2, batch_shape=[3])<tf.Tensor: shape=(3, 2, 2), dtype=float32, numpy=array([[[1., 0.], [0., 1.]], [[1., 0.],

import tensorflow as tfimage = tf.zeros([10,10,3])print(image.shape.as_list())print(tf.expand_dims(image, axis=0).shape.as_list())print(tf.expand_dims(image, axis=1).shape.as_list())cc=tf.expand_dims(image, -1)print(cc.shape.as_list())print(

import osimport tensorflow as tfimport cProfiletf.executing_eagerly() x = [[2.]]m = tf.matmul(x, x)print("hello, {}".format(m))hello, [[4.]]

x = tf.constant([1,2,3])print(x.shape)x = tf.ensure_shape(x, [3])

import tensorflow as tf # 'hypothesis' is a tensor of shape `[2, 1]` with variable-length values:# (0,0) = ["a"]# (1,0) = ["b"]hypothesis = tf.sparse.SparseTensor( [[0, 0, 0], [1, 0, 0]], ["a", "b"], (2, 1, 1))tf.sparse.to_dense(

为什么用tf.sparse.SparseTensor的原因节约内存在表达一个稀疏矩阵时候，不用一个很大的矩阵tf.sparse.SparseTensor( indices, values, dense_shape)indices ：非零值的元素的索引values ：非零值dense_shape ：shapeimport tensorflow as tfsp_input=tf.sparse.SparseTensor(indices=[[0, 0], [

import tensorflow as tfx=tf.constant([0.1, -1., 5.2, 4.3, -1., 7.4])#判断x里面的元素是否是１condition_mask=tf.not_equal(x,tf.constant(-1.))#[ True, False, True, True, False, True]#将张量拆成两个,按照condition_mask的对应位置partitioned_data = tf.dynamic_partition(

import tensorflow as tf import numpy as npdef my_func(arg): arg = tf.convert_to_tensor(arg, dtype=tf.float32) return arg# The following calls are equivalent.value_1 = my_func(tf.constant([[1.0, 2.0], [3.0, 4.0]]))print(value_1)value_2 = my_.

import tensorflow as tfdef make_variables(k, initializer): return (tf.Variable(initializer(shape=[k], dtype=tf.float32)), tf.Variable(initializer(shape=[k, k], dtype=tf.float32)))v1, v2 = make_variables(3, tf.constant_initializer(2.))v1Out

import tensorflow as tfx = tf.constant(12)y = tf.constant(5)def f1(): return tf.multiply(x, 17)def f2(): return tf.add(y, 23)r = tf.cond(tf.less(x, y), f1, f2)print(r.numpy())如果 x<y，就执行函数 f1 , 否则 就执行 f228

import tensorflow as tft1 = [[1, 2, 3], [4, 5, 6]]t2 = [[7, 8, 9], [10, 11, 12]]tf.concat([t1, t2], 0)<tf.Tensor: shape=(4, 3), dtype=int32, numpy=array([[ 1, 2, 3], [ 4, 5, 6], [ 7, 8, 9], [10, 11, 12]], dtype=int32)&gt

t = tf.constant([[-10., -1., 0.], [0., 2., 10.]])t2 = tf.clip_by_value(t, clip_value_min=-1, clip_value_max=1)t2.numpy()array([[-1., -1., 0.], [ 0., 1., 1.]], dtype=float32)

t= [ x1x_{1}x1​, x2x_{2}x2​,… x2x_{2}x2​]l2=∑i=0nxi2\sqrt{\displaystyle\sum\limits_{i=0}^n x_i^2}i=0∑n​xi2​​norm = 2.0clip_norm=t∗norml2=\frac{t*norm}{l2}=l2t∗norm​import numpy as npt=np.array([[1, 2, 3, 4, 5]])l2norm4t = np.linalg.norm(t)clip_