首先咱們分析一下下面的代碼:dom
import tensorflow as tf import numpy as np a=tf.constant([[1., 2., 3.],[4., 5., 6.]]) b=np.float32(np.random.randn(3,2)) #c=tf.matmul(a,b) c=tf.multiply(a,b) init=tf.global_variables_initializer() with tf.Session() as sess: print(c.eval())
問題是上面的代碼編譯正確嗎?編譯一下就知道,錯誤信息以下:函數
ValueError: Dimensions must be equal, but are 2 and 3 for 'Mul' (op: 'Mul') with input shapes: [2,3], [3,2].spa
顯然,tf.multiply()表示點積,所以維度要同樣。而tf.matmul()表示普通的矩陣乘法。code
並且tf.multiply(a,b)和tf.matmul(a,b)都要求a和b的類型必須一致。可是之間存在着細微的區別。對象
在tf中全部返回的tensor,無論傳進去是什麼類型,傳出來的都是numpy ndarray對象。blog
看看官網API介紹:ip
tf.matmul( a, b, transpose_a=False, transpose_b=False, adjoint_a=False, adjoint_b=False, a_is_sparse=False, b_is_sparse=False, name=None ) tf.multiply( x, y, name=None )
可是tf.matmul(a,b)函數不只要求a和b的類型必須徹底一致,同時返回的tensor類型同a和b一致;而tf.multiply(a,b)函數僅要求a和b的類型顯式一致,同時返回的tensor類型與a一致,即在不聲明類型的狀況下,編譯不報錯。input
例如:it
#類型一致,能夠運行 import tensorflow as tf import numpy as np a=tf.constant([[1, 2, 3],[4, 5, 6]],dtype=np.float32) b=np.float32(np.random.randn(3,2)) c=tf.matmul(a,b) #c=tf.multiply(a,b) init=tf.global_variables_initializer() with tf.Session() as sess: print (type(c.eval()),type(a.eval()),type(b))
#類型不一致,不能夠運行 import tensorflow as tf import numpy as np a=tf.constant([[1, 2, 3],[4, 5, 6]]) b=np.float32(np.random.randn(3,2)) c=tf.matmul(a,b) #c=tf.multiply(a,b) init=tf.global_variables_initializer() with tf.Session() as sess: print (type(c.eval()),type(a.eval()),type(b))
#類型不一致,能夠運行,結果的類型和a一致 import tensorflow as tf import numpy as np a=tf.constant([[1, 2, 3],[4, 5, 6]]) b=np.float32(np.random.randn(2,3)) #c=tf.matmul(a,b) c=tf.multiply(a,b) init=tf.global_variables_initializer() with tf.Session() as sess: print (c.eval()) print (type(c.eval()),type(a.eval()),type(b))
#類型不一致,不能夠運行 import tensorflow as tf import numpy as np a=tf.constant([[1, 2, 3],[4, 5, 6]], dtype=np.float32) b=tf.constant([[1, 2, 3],[4, 5, 6]], dtype=np.int32) #c=tf.matmul(a,b) c=tf.multiply(a,b) init=tf.global_variables_initializer() with tf.Session() as sess: print (c.eval()) print (type(c.eval()),type(a.eval()),type(b))