Question

有人可以解释为什么在运行以下代码时tensorflow给我带来麻烦吗？

import tensorflow as tf

x = tf.keras.layers.Input(shape=(1,))
y = tf.keras.layers.Dense(1, activation=tf.nn.relu)(x)

loss = tf.losses.mean_squared_error(x,y)
grad = tf.gradients(loss, tf.trainable_variables())

# !!! GIVES ME TROUBLE !!!
clipped_grad = tf.clip_by_norm(grad, 1.0)

with tf.Session() as sess:
    sess.run(tf.global_variables_initializer())
    sess.run(y, feed_dict={x: [[1.0], [2.0], [3.0]]})

我得到的错误：

Traceback (most recent call last):
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 1589, in _create_c_op
    c_op = c_api.TF_FinishOperation(op_desc)
tensorflow.python.framework.errors_impl.InvalidArgumentError: Shapes must be equal rank, but are 2 and 1
    From merging shape 0 with other shapes. for 'clip_by_norm/t' (op: 'Pack') with input shapes: [1,1], [1].

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:/Users/TObs/.PyCharmCE2018.1/config/scratches/scratch.py", line 11, in <module>
    clipped_grad = tf.clip_by_norm(grad, 1.0)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\ops\clip_ops.py", line 140, in clip_by_norm
    t = ops.convert_to_tensor(t, name="t")
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 1011, in convert_to_tensor
    as_ref=False)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 1107, in internal_convert_to_tensor
    ret = conversion_func(value, dtype=dtype, name=name, as_ref=as_ref)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\ops\array_ops.py", line 960, in _autopacking_conversion_function
    return _autopacking_helper(v, inferred_dtype, name or "packed")
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\ops\array_ops.py", line 923, in _autopacking_helper
    return gen_array_ops.pack(elems_as_tensors, name=scope)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\ops\gen_array_ops.py", line 5532, in pack
    "Pack", values=values, axis=axis, name=name)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\op_def_library.py", line 787, in _apply_op_helper
    op_def=op_def)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 3414, in create_op
    op_def=op_def)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 1756, in __init__
    control_input_ops)
  File "D:\Program Files\Python\Python_3_6_2\lib\site-packages\tensorflow\python\framework\ops.py", line 1592, in _create_c_op
    raise ValueError(str(e))
ValueError: Shapes must be equal rank, but are 2 and 1
    From merging shape 0 with other shapes. for 'clip_by_norm/t' (op: 'Pack') with input shapes: [1,1], [1].

有什么想法吗？我正在使用tensorflow-gpu 1.9.0，NVidia GTX 1080的Windows10计算机上运行。

我们将不胜感激：）

干杯，滚滚。

Answer 1

因此，在修改之后，我发现必须分别对渐变张量liek中的每个值应用tf.clip_by_norm，这样：

clipped_gradients = [tf.clip_by_norm（g，grad_norm_clip）for tf.gradients（loss，tf.trainable_variables（））中的g

我猜，那是正确的方法，对吧？

干杯，滚滚。

tf.clip_by_norm（grad，1.0）引发InvalidArgumentError形状必须等于等级，但必须为2和1

1 个答案: