Question

a = tf.compat.v1.constant(5.0)
b = tf.compat.v1.constant(6.0)

sum1 = a + b
g = tf.compat.v1.Graph()
with g.as_default():
    # Define operations and tensors in `g`.
    hello = tf.compat.v1.constant('hello')
    assert hello.graph is g

sess = tf.compat.v1.Session(graph=g)

print(sess.run(sum1))

tensorflow-gpu2.0 我不知道为什么。我是tensorflow的初学者

Answer 1

导入 tensorflow 后需要禁用 Eager Execution，如下所示：

import tensorflow as tf
tf.compat.v1.disable_eager_execution()

它对我有用。

Answer 2

我不知道你想做什么！但我做了一些猜测，这就是我得出的结果。

import tensorflow as tf

g = tf.compat.v1.Graph()
with g.as_default():
    a = tf.compat.v1.constant(5.0)
    b = tf.compat.v1.constant(6.0)
    sum1 = tf.add(a, b) # instead of sum1 = a + b
    
    hello = tf.compat.v1.constant('hello')
    assert hello.graph is g

sess = tf.compat.v1.Session(graph=g)

print(sess.run(sum1))

我相信它适用于 Tensorflow 2.x，如果您仍然遇到急切的异常，那么您应该添加：

tf.compat.v1.disable_eager_execution()

之前

assert hello.graph is g

Answer 3

我尝试了Tensorflow版本1.14.0

执行以下操作

select member,
       min(case when prev_tot = 'PREVENTATIVE' and tot = 'TREATMENT'
                then date
           end) as first_prev_to_treat_date
from (select t.*,
             lag(tot) over (partition by member order by date) as prev_tot
      from t
     )  t
group by member;

尝试以下代码

pip install tensorflow==1.14..0

启用急切执行时，Tensor.graph没有意义

3 个答案: