Clever Geek Handbook

How to get scales from tensorflow fully_linked

I am trying to extract weights from a model after training. Here is an example of a toy

import tensorflow as tf
import numpy as np

X_ = tf.placeholder(tf.float64, [None, 5], name="Input")
Y_ = tf.placeholder(tf.float64, [None, 1], name="Output")

X = ...
Y = ...
with tf.name_scope("LogReg"):
    pred = fully_connected(X_, 1, activation_fn=tf.nn.sigmoid)
    loss = tf.losses.mean_squared_error(labels=Y_, predictions=pred)
    training_ops = tf.train.GradientDescentOptimizer(0.01).minimize(loss)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    for i in range(200):
        sess.run(training_ops, feed_dict={
            X_: X,
            Y_: Y
        })
        if (i + 1) % 100 == 0:
            print("Accuracy: ", sess.run(accuracy, feed_dict={
                X_: X,
                Y_: Y
            }))

# Get weights of *pred* here

I have looked at Get weights from tensorflow model and docs but can't find a way to get the weights value.

So, in the toy example, suppose X_ has shape (1000, 5), how can I get 5 values ​​in weights of 1 layer after

+3


source to share


2 answers


There are some issues in your code that need to be fixed:

1- You need to use variable_scope

instead name_scope

in the following line (see TensorFlow documentation for the difference between the two):

with tf.name_scope("LogReg"):

2- To be able to get a variable later in the code, you need to know its name. So, you need to assign a name to the variable of interest (if you don't support it, it will be assigned by default, but then you need to figure out what it is!):



pred = tf.contrib.layers.fully_connected(X_, 1, activation_fn=tf.nn.sigmoid, scope = 'fc1')

Now let's see how the above fixes can help us get the value of a variable. Each layer has two types of variables: weights and biases. In the following code snippet (your modified version), I will only show you how to extract the weight for a fully connected layer:

X_ = tf.placeholder(tf.float64, [None, 5], name="Input")
Y_ = tf.placeholder(tf.float64, [None, 1], name="Output")

X = np.random.randint(1,10,[10,5])
Y = np.random.randint(0,2,[10,1])

with tf.variable_scope("LogReg"):
    pred = tf.fully_connected(X_, 1, activation_fn=tf.nn.sigmoid, scope = 'fc1')
    loss = tf.losses.mean_squared_error(labels=Y_, predictions=pred)
    training_ops = tf.train.GradientDescentOptimizer(0.01).minimize(loss)

with tf.Session() as sess:

    all_vars= tf.global_variables()
    def get_var(name):
        for i in range(len(all_vars)):
            if all_vars[i].name.startswith(name):
                return all_vars[i]
        return None
    fc1_var = get_var('LogReg/fc1/weights')

    sess.run(tf.global_variables_initializer())    
    for i in range(200):
        _,fc1_var_np = sess.run([training_ops,fc1_var], feed_dict={
        X_: X,
        Y_: Y 
        })
        print fc1_var_np
+8


source


Try the following:

with tf.Session() as sess:
    last_check = tf.train.latest_checkpoint(tf_data)
    saver = tf.train.import_meta_graph(last_check+'.meta')
    saver.restore(sess,last_check)
    ######
    Model_variables = tf.GraphKeys.MODEL_VARIABLES
    Global_Variables = tf.GraphKeys.GLOBAL_VARIABLES
    ######
    all_vars = tf.get_collection(Model_variables)
    # print (all_vars)
    for i in all_vars:
        print (str(i) + '  -->  '+ str(i.eval()))


I got it:

<tf.Variable 'linear/linear_model/DOLocationID/weights/part_0:0' shape=(1, 1) dtype=float32_ref>  -->  [[-0.00912262]]
<tf.Variable 'linear/linear_model/PULocationID/weights/part_0:0' shape=(1, 1) dtype=float32_ref>  -->  [[ 0.00573495]]
<tf.Variable 'linear/linear_model/passenger_count/weights/part_0:0' shape=(1, 1) dtype=float32_ref>  -->  [[-0.07072949]]
<tf.Variable 'linear/linear_model/trip_distance/weights/part_0:0' shape=(1, 1) dtype=float32_ref>  -->  [[ 2.59973669]]
<tf.Variable 'linear/linear_model/bias_weights/part_0:0' shape=(1,) dtype=float32_ref>  -->  [ 4.27982235]
0


source






More articles:


All Articles