search

brain-research / realistic-ssl-evaluation

Open source release of the evaluation benchmark suite described in "Realistic Evaluation of Deep Semi-Supervised Learning Algorithms"
Apache License 2.0
458 stars 98 forks source link

Mark global_step_init as used to prevent spurious warning msg #14

Open gar1t opened 5 years ago

gar1t commented 5 years ago

Without this, TF displays a warning message that looks like an error (contains traceback):

E1121 12:34:00.888499 139860141586240 tf_logging.py:105] ==================================
Object was never used (type <class 'tensorflow.python.framework.ops.Operation'>):
<tf.Operation 'init' type=NoOp>
If you want to mark it as used call its "mark_used()" method.
It was originally created here:
  ...

This behavior occurs on TF 1.12.0 but reports suggest that these warnings started appearing as early as TF 1.2.0.

craffel commented 5 years ago

Hmm, I think a better solution is for global_step_init to be created in train_model.py instead. What do you think?