TensorFlow 기초 2 - 변수, 값 치환, 누적(+=)

 

 

# 변수 : 모델 학습 시, 매개변수 갱신 등을 위해 사용

import tensorflow as tf

f = tf.Variable(1.0)
v = tf.Variable(tf.ones((2,)))
m = tf.Variable(tf.ones((2, 1)))
print(f)
print(v)
print(m)
print(m.numpy())

print()
v1 = tf.Variable(1)
print(v1)
v1.assign(10) # 값 치환
print(v1, v1.numpy(), type(v1))

v2 = tf.Variable(tf.ones(shape=(1))) # 1-d tensor
v2.assign([20])
print(v2, v2.numpy(), type(v2))

v3 = tf.Variable(tf.ones(shape=(1, 2))) # 2-d tensor
v3.assign([[20, 30]])
print(v3, v3.numpy(), type(v3))

print()
v1 = tf.Variable([3])
v2 = tf.Variable([5])
v3 = v1 * v2 + 10
print(v3)

print()
var = tf.Variable([1,2,3,4,5], dtype = tf.float32)
result1 = var + 10
print(result1)

print()
w = tf.Variable(tf.ones(shape=(1,)))
b = tf.Variable(tf.ones(shape=(1,)))
w.assign([2])
b.assign([3])

def func1(x):
    return w * x + b

out_a1 = func1([[3]])  # x값에 의해서 리턴값의 차원이 달라질 수 있다.
print('out_a1 :', out_a1)

print()
@tf.function # auto graph 기능이 적용된 함수 : tf.Graph + tf.Session이 적용
def func2(x):
    return w * x + b

print(type(func2))
out_a2 = func2([1, 2])
print('out_a1 :', out_a2)

print()
rand = tf.random.uniform([1], 0, 1) # 균등분포
print(rand.numpy())
rand2 = tf.random.normal([5], mean=0, stddev=1) # 정규분포
print(rand2.numpy())

print()
aa = tf.ones((2, 1))
print(aa.numpy())

m = tf.Variable(tf.zeros((2, 1)))
print(m.numpy())
m.assign(aa)
print(m.numpy())

m.assign_add(aa)  # m += aa의 개념
print(m.numpy())

m.assign_sub(aa)  # m -= aa의 개념
print(m.numpy())



<console>
2022-11-28 14:41:27.387413: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
<tf.Variable 'Variable:0' shape=() dtype=float32, numpy=1.0>
<tf.Variable 'Variable:0' shape=(2,) dtype=float32, numpy=array([1., 1.], dtype=float32)>
<tf.Variable 'Variable:0' shape=(2, 1) dtype=float32, numpy=
array([[1.],
       [1.]], dtype=float32)>
[[1.]
 [1.]]

<tf.Variable 'Variable:0' shape=() dtype=int32, numpy=1>
<tf.Variable 'Variable:0' shape=() dtype=int32, numpy=10> 10 <class 'tensorflow.python.ops.resource_variable_ops.ResourceVariable'>
<tf.Variable 'Variable:0' shape=(1,) dtype=float32, numpy=array([20.], dtype=float32)> [20.] <class 'tensorflow.python.ops.resource_variable_ops.ResourceVariable'>
<tf.Variable 'Variable:0' shape=(1, 2) dtype=float32, numpy=array([[20., 30.]], dtype=float32)> [[20. 30.]] <class 'tensorflow.python.ops.resource_variable_ops.ResourceVariable'>

tf.Tensor([25], shape=(1,), dtype=int32)

tf.Tensor([11. 12. 13. 14. 15.], shape=(5,), dtype=float32)

out_a1 : tf.Tensor([[9.]], shape=(1, 1), dtype=float32)

<class 'tensorflow.python.eager.polymorphic_function.polymorphic_function.Function'>
out_a1 : tf.Tensor([5. 7.], shape=(2,), dtype=float32)

[0.35073256]
[-2.0095115 -0.3110462 -1.9975616  0.3110392  0.5237793]

[[1.]
 [1.]]
[[0.]
 [0.]]
[[1.]
 [1.]]
[[2.]
 [2.]]
[[1.]
 [1.]]

 

Variable() 함수는 값을 치환할 경우 assign() 함수를 사용하여 친환하여야 된다. 평범하게 변수를 치환하면 에러를 발생시킨다.

 

tensor는 tensor끼리 연산해야 더 빠르게 연산된다.