Advertisement
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- c = tf.ragged.constant([[1, 2, 3], [4, 5]])
- v = tf.ragged.constant([[10., 20., 30.], [40., 50.]])
- r = tf.random.uniform([1, 1], maxval=2, dtype=tf.int32)
- with tf.Session() as sess:
- print(sess.run([tf.gather_nd(c, r), tf.gather_nd(v, r)]))
- import tensorflow as tf
- tf.enable_eager_execution() # you can use a normal Session, but this to show intermediate output
- c = tf.ragged.constant([[1, 2, 3], [4, 5]])
- v = tf.ragged.constant([[10., 20., 30.], [40., 50.]])
- r = tf.random.uniform([1, 1], maxval=2, dtype=tf.int32)
- a = tf.gather_nd(c, r)
- b = tf.gather_nd(v, r)
- print(a)
- print(b)
- # Output example
- #<tf.RaggedTensor [[1, 2, 3]]>
- #<tf.RaggedTensor [[10.0, 20.0, 30.0]]>
- # Lengths
- l_a = tf.squeeze(a.row_lengths())
- l_b = tf.squeeze(b.row_lengths())
- print(l_a)
- print(l_b)
- #Output example
- #tf.Tensor(3, shape=(), dtype=int64)
- #tf.Tensor(3, shape=(), dtype=int64)
- #Random index between 0 and length
- rand_idx_a = tf.random.uniform([1],minval=0,maxval=l_a,dtype=tf.int64)
- rand_idx_b = tf.random.uniform([1],minval=0,maxval=l_b,dtype=tf.int64)
- print(rand_idx_a)
- print(rand_idx_b)
- #Output example
- #tf.Tensor([0], shape=(1,), dtype=int64)
- #tf.Tensor([2], shape=(1,), dtype=int64)
- #Convert ragged tensor to tensor of shape [1,n]
- t_a = a.to_tensor()
- t_b = b.to_tensor()
- print(t_a)
- print(t_b)
- #Read from extracted tensor using random index
- rand_a = tf.gather_nd(tf.squeeze(t_a),rand_idx_a) #removes dimension of 1
- rand_b = tf.gather_nd(tf.squeeze(t_b),rand_idx_b)
- print(rand_a)
- print(rand_b)
- #Output example
- #tf.Tensor([[1 2 3]], shape=(1, 3), dtype=int32)
- #tf.Tensor([[10. 20. 30.]], shape=(1, 3), dtype=float32)
- #tf.Tensor(1, shape=(), dtype=int32)
- #tf.Tensor(30.0, shape=(), dtype=float32)
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement