Currently, we are trying to implement sampling using tensorflow.
I'd like to save the results by sampling several times, but I'm having trouble with the following feeling.
Here's a quick example.
x=tf.Variable(0)
step=tf.assign_add(x,1)
Prepare the above (absolutely use the above) and try to get y=[123] by repeating step 3 times.
y=[step for_in range(3)]
As , y returns as [111].
Is it possible to run the steps several times in order (not at the same time)?
x=tf.Variable(0)
y = tf.Variable([0]*3)
def fun(i):
step=tf.assign_add(x,1)
assign=tf.assign(y[i], step)
with tf.control_dependencies ([assign]):
return i+1
with tf.Session() asess:
tf.global_variables_initializer().run()
result=tf.while_loop(lambdai:i<3,fun,[0])
sess.run(result)
print(y.eval())
Now I can do what I thought.
© 2024 OneMinuteCode. All rights reserved.