我的学习代码有些问题，因为它在执行时会返回以下错误
错误
Traceback (most recent call last):
File "tf_apprentissage/main.py", line 109, in <module>
f_de_x = model(x) # permet de déterminer ce qu'est l'image (y = f(x) )
File "tf_apprentissage/main.py", line 91, in model
c9 = dense(c7, (-1,100, 2))
File "tf_apprentissage/main.py", line 61, in dense
w = tf.Variable(tf.truncated_normal(forme, stddev=(2/n)**.5)) # poid
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/ops/random_ops.py", line 169, in truncated_normal
stddev_tensor = ops.convert_to_tensor(stddev, dtype=dtype, name="stddev")
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/framework/ops.py", line 836, in convert_to_tensor
as_ref=False)
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site- packages/tensorflow/python/framework/ops.py", line 926, in internal_convert_to_tensor
ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/framework/constant_op.py", line 229, in _constant_tensor_conversion_function
return constant(v, dtype=dtype, name=name)
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/framework/constant_op.py", line 208, in constant
value, dtype=dtype, shape=shape, verify_shape=verify_shape))
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/framework/tensor_util.py", line 383, in make_tensor_proto
_AssertCompatible(values, dtype)
File "/home/etudiant/PycharmProjects/tf_apprentissage/lib/python3.5/site-packages/tensorflow/python/framework/tensor_util.py", line 303, in _AssertCompatible
(dtype.name, repr(mismatch), type(mismatch).__name__))
TypeError: Expected float32, got (8.659560562354934e-17+1.4142135623730951j) of type 'complex' instead.
我是Tensorflow的初学者，所以我不知道问题出在哪里，我会非常感谢任何帮助。
此源代码的目标是识别类项目的单词，首先我们要识别字母和数字。我之前从未做过任何机器学习项目，这是我第一次使用tensorflow。我们使用绘制数字的图像作为带有标签的输入。
我的代码是
# -*- coding: utf-8 -*-
import tensorflow as tf
from PIL import Image
import numpy as np
import os
"""
# Model parameters
W = tf.Variable([.3], dtype=tf.float32)
b = tf.Variable([-.3], dtype=tf.float32)
# Model input and output
x = tf.placeholder(tf.float32)
linear_model = W * x + b
y = tf.placeholder(tf.float32)
# loss
loss = tf.reduce_sum(tf.square(linear_model - y)) # sum of the squares
# optimizer
optimizer = tf.train.GradientDescentOptimizer(0.01)
train = optimizer.minimize(loss)
# training data
x_train = [1, 2, 3, 4]
y_train = [0, -1, -2, -3]
# training loop
init = tf.global_variables_initializer()
sess = tf.Session()
sess.run(init) # reset values to wrong
for i in range(1000):
sess.run(train, {x: x_train, y: y_train})
# evaluate training accuracy
curr_W, curr_b, curr_loss = sess.run([W, b, loss], {x: x_train, y: y_train})
print("W: %s b: %s loss: %s"%(curr_W, curr_b, curr_loss))
"""
def conv(entree, noyaux):
n = noyaux[0]*noyaux[1]*noyaux[2]
W = tf.Variable(tf.truncated_normal(noyaux, stddev=(2/n)**.5)) # poid aléatoire (distribution gaussienne)
b = tf.Variable(tf.zeros(noyaux[-1:])) # biais
conv = tf.nn.conv2d(entree, W, [1, 1, 1, 1], 'SAME')+b
# entree = notre image ,
# W = poid (noyaux de notre convolution,
# [1,1,1,1] = pas de notre convolution,
# 'SAME' va rajouter des zeros autour de notre matrice permet une matrice de même taille
# donc en sortie pour que le noyaux ne perde pas d'information
return tf.nn.relu(conv) # on applique la non linéarité
def pool(entree):
return tf.nn.max_pool(
entree,
[1, 2, 2, 1],
[1, 2, 2, 1],
'SAME'
)
def dense(entree, forme):
n = forme[0]
w = tf.Variable(tf.truncated_normal(forme, stddev=(2/n)**.5)) # poid
b = tf.Variable(tf.zeros(forme[0]))
matmul = tf.matmul(entree, w)+b
return tf.nn.relu(matmul)
def model(entree):
c1 = conv(entree, (3, 3, 3, 16))
c2 = conv(c1, (3, 3, 16, 16))
c3 = pool(c2)
c4 = conv(c3, (3, 3, 16, 32))
# explication du deuxième paramètre
# 3 -> Longueur
# 3 -> Largeur
# 16->profondeur de la matrice 3D précédente
# 32->profondeur de la prochaine matrice 3D
c5 = conv(c4, (3, 3, 32, 32))
c6 = pool(c5)
c7 = tf.reshape(c6, (-1, 16**2 * 32)) # on ecrase notre matrice en 3d dans un vecteur
# c6 = la matrice 3D
# arg2 = la forme :
# -1 : qu'une solution
# 16**2 * 32 = nombre d'element dans le vecteur
c8 = dense(c7, (-1,16**2 * 32, 100)) # multiplication matricielle
c9 = dense(c8, (-1,100, 2))
return c9
def resize_img(img):
image_temp = Image.open(img)
image = tf.image.decode_png(content = image_temp, channels = 1, dtype = tf.float32)
image = tf.image.resize_images(image_temp,[64,64])
return image_temp
if __name__ == "__main__":
x = tf.placeholder(tf.float32, [None, 64, 64, 3]) # contiendra nos images par la suite
# tf.float32 -> type de la donnée qui serra stocké
# argument 2 vas être la forme de x
# None -> signifie qu'on traite plusieurs image en même temps
# 64, 64, 3 -> résolution de l'image : 3 = profondeur
path_to_dataset = "Hnd/Img/"
f_de_x = model(x) # permet de déterminer ce qu'est l'image (y = f(x) )
y = tf.placeholder(tf.float32, [None, 2]) # les labels (ce que signifie en réalité l'image
# fonction de cout (calcul de la distance entre ce que l'on veut et ce qu'on a
cout = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(logits=f_de_x, labels=y))
step = tf.train.AdamOptimizer(0.0001).minimize(cout) # on minimise notre cout (0.0001 est arbitraire)
with tf.Session() as session:
session.run(tf.global_variables_initializer())
for dir in os.listdir(path_to_dataset):
for file in os.listdir(path_to_dataset + dir):
image = resize_img(path_to_dataset+dir+"/"+file)
session.run(step, feed_dict={x: image, y: dir})
答案 0 :(得分：0)
w = tf.Variable(tf.truncated_normal(forme, stddev=(2/n)**.5)) # poid
forme似乎是一个复数，在您引用forme[0]的上方和下方的行上，我认为这是复数的真实部分。而在上面显示的行上forme被引用而没有索引到数字的实部。
你想要：
w = tf.Variable(tf.truncated_normal(forme[0], stddev=(2/n)**.5)) # poid
如果没有，请打印出forme的值并验证它是否符合您的预期。如果你并不意味着它是一个复杂的数字，那就追查那个问题吧。