Global step not increased

DeepSpeech
#1

I’m getting the following warning while training using pre-trained weights with CV data.

It seems that global step (tf.train.get_global_step) has not been increased. Current value (could be stable): 269108 vs previous value: 269108. You could increase the global step by passing tf.train.get_global_step() to Optimizer.apply_gradients or Optimizer.min

Can someone please let me know how to fix this? The model seems to do bad upon training.

#2

Have you tried doing what the warning tells you to do?

#3

@reuben
How about this? i still haven’t checked if it works because i don’t want to interrupt training just yet.

Line 1519 DeepSpeech.py

`
apply_gradient_op = optimizer.apply_gradients(avg_tower_gradients, global_step=tf.train.get_global_step())
`

am i passing it the right way? or shall i just do the following instead:

`
    apply_gradient_op = optimizer.apply_gradients(avg_tower_gradients, tf.train.get_global_step())

`