change n_steps value in training and test about tensorflow-examples HOT 4 CLOSED

What you are looking for is RNNs with dynamic calculation, you need to change rnn function and set a 'sequence_length' argument that will represents a list of each sequence length in your batch.
For more info: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/ops/rnn.py#57

from tensorflow-examples.

Hi,

I used the code with _seq_len. Actually in training, n_steps (n_steps is global variable) is set constantly as 16, so as in your code there will be a vector _seq_len, and split with n_steps:

_seq_len = tf.fill(_batch_size, constant(_seq_len, dtype=tf.int32)) _seq_len = tf.cast(_seq_len, dtype=tf.int64) #seq_len needs type int64.. ... ... _X = tf.split(0,n_steps, _X) # n_steps * (batch_size, n_hidden)

but when testing, how can i change this n_steps value into 55 in tf.split? does it make sense?

from tensorflow-examples.

my problem maybe less complicated, because when training n_steps is constant, What I want to do is just change n_steps values when test

from tensorflow-examples.

hello, Did your problem solved?

from tensorflow-examples.