tensorflow檢查op是否可導(反向傳播)
1.安裝最新版tf--tensorflow1.5,gpu版本需要CUDA8和cudnn6,命令如下
GPU版:sudo pip3 install tf-nightly-gpu
CPU版:sudo pip3 install tf-nightly
對應pip網站:https://pypi.python.org/pypi/tf-nightly-gpu
2.編寫程式碼進行測試,主要包括可導函式square和不可導函式floor
程式碼參考網站https://research.googleblog.com/2017/10/eager-execution-imperative-define-by.html
程式碼示例:
import numpy as np
import tensorflow as tf
import tensorflow.contrib.eager as tfe
tfe.enable_eager_execution()
def floor(x):
return tf.floor(x)
def square(x):
return tf.multiply(x, x)
grad_f = tfe.gradients_function(floor)
print(floor(3.))
print('gradient of floor:',grad_f(3.))
grad_s = tfe.gradients_function(square)
print(square(3.))
print('gradient of square:',grad_s([3.]))
程式碼輸出:
tf.Tensor(3.0, shape=(), dtype=float32)
gradient of floor: [None]
tf.Tensor(9.0, shape=(), dtype=float32)
gradient of square: [<tf.Tensor: id=21, shape=(1,), dtype=float32, numpy=array([ 6.], dtype=float32)>]
3.小結
可以看出,floor函式對應的梯度為None,而square函式對應的梯度為 derivative(x^2)=2*x|x=3=6
福利區:獨學而無友,則孤陋而寡聞.加入機器學習與深度學習討論QQ群(581789266),一塊交流與探討,共同成長和進步!