Question

from sklearn.datasets import fetch_california_housing
import numpy as np
from sklearn.preprocessing import StandardScaler


n_epochs = 1000
learning_rate = 0.01

housing = fetch_california_housing()

# m is length of dataset, n is # of atts
m, n = housing.data.shape

# Scale Data with StandardScalar()
scaler = StandardScaler()
scaled_housing_data = scaler.fit_transform(housing.data)

# m is length of dataset, n is # of atts
m, n = housing.data.shape

batch_size = 100
num_batches = int(np.ceil(m/batch_size))


# Just appends a a column of '1.'s to beginnging of table
scaled_housing_data_plus_bias = np.c_[np.ones((m, 1)), scaled_housing_data]

X = tf.placeholder(tf.float32, shape=(None, n+1), name="X")
y = tf.placeholder(tf.float32, shape=(None, 1), name="y")
theta = tf.Variable(tf.random_uniform([n+1, 1], -1, 1), name="theta")

# Formulas for preducing predictions and error metrics
y_pred = tf.matmul(X, theta)
error = y_pred - y
mse = tf.reduce_mean(tf.square(error), name="mse")


# gradients = 2/m * tf.matmul(tf.transpose(X), error)
gradients = tf.gradients(mse, [theta])[0]       # Calculate gradients with built in function for efficiency

# Operation that adjusts theta values based on gradients & learning rate
# training_op = theta.assign(theta - learning_rate * gradients)
# optimizer = tf.train.GradientDescentOptimizer(learning_rate=learning_rate)
optimizer = tf.train.MomentumOptimizer(learning_rate=learning_rate, momentum=0.9)
training_op = optimizer.minimize(mse)



def fetch_batch(epoch, batch_index, batch_size):
    np.random.seed(epoch * num_batches + batch_index)  
    indices = np.random.randint(m, size=batch_size)
    X_batch = scaled_housing_data_plus_bias[indices]
    y_batch = housing.target.reshape(-1, 1)[indices]


    return X_batch, y_batch

init = tf.global_variables_initializer()

with tf.Session() as sess:
    sess.run(init)

    for epoch in range(n_epochs):
        print("Epoch: " + str(epoch))

        for batch_index in range(num_batches):
            X_batch, y_batch = fetch_batch(epoch, batch_index, batch_size)
            print(X_batch.shape)
            print(y_batch.shape)
            sess.run(training_op, feed_dict={X: X_batch, y: y_batch})
            if epoch % 100 == 0:
                print("Mean Squared Error: " + str(mse.eval()))

    print(theta.eval())

输出：纪元：0，（100,9），（100,1）

Error Message: 
InvalidArgumentError (see above for traceback): You must feed a value for placeholder tensor 'y' with dtype float and shape [?,1]
     [[Node: y = Placeholder[dtype=DT_FLOAT, shape=[?,1], _device="/job:localhost/replica:0/task:0/device:CPU:0"]()]]

（我也得到类似'X'的东西）提前感谢任何建议或帮助!!!

Answer 1

我明白了。尝试评估'mse'会导致占位符值出错。

# Correct Code
with tf.Session() as sess:
    sess.run(init)
    best_theta = None
    best_mse = None
        for epoch in range(n_epochs):

        for batch_index in range(num_batches):
            X_batch, y_batch = fetch_batch(epoch, batch_index, batch_size)
            _, best_mse = sess.run([training_op,mse], feed_dict={X: X_batch, y: y_batch})

        if epoch % 100 == 0:
            print("Epoch: " + str(epoch))
            print("MSE " + str(best_mse))

     print(theta.eval())

Tensorflow - 即使形状看起来没问题，也会在feed_dict（）上出现形状错误

1 个答案: