Upload fail: Module 'tensorflow._api.v1.io' has no attribute 'gfile'

#1

Hi,
We had successful uploads and evaluation recently, but since our last update, we had this error which resulted in upload failure:

2019-07-13T08:28:07.810465863Z Traceback (most recent call last):
2019-07-13T08:28:07.810526387Z   File "/srv/conda/envs/notebook/lib/python3.6/runpy.py", line 193, in _run_module_as_main
2019-07-13T08:28:07.810752237Z     "__main__", mod_spec)
2019-07-13T08:28:07.810791615Z   File "/srv/conda/envs/notebook/lib/python3.6/runpy.py", line 85, in _run_code
2019-07-13T08:28:07.810943842Z     exec(code, run_globals)
2019-07-13T08:28:07.810960772Z   File "/home/aicrowd/dopamine/discrete_domains/train.py", line 28, in <module>
2019-07-13T08:28:07.811271198Z     from dopamine.discrete_domains import run_experiment
2019-07-13T08:28:07.811290649Z   File "/home/aicrowd/dopamine/discrete_domains/run_experiment.py", line 25, in <module>
2019-07-13T08:28:07.811408312Z     from dopamine.agents.dqn import dqn_agent
2019-07-13T08:28:07.811420456Z   File "/home/aicrowd/dopamine/agents/dqn/dqn_agent.py", line 28, in <module>
2019-07-13T08:28:07.811561674Z     from dopamine.discrete_domains import atari_lib
2019-07-13T08:28:07.81157627Z   File "/home/aicrowd/dopamine/discrete_domains/atari_lib.py", line 37, in <module>
2019-07-13T08:28:07.811651847Z     import gin.tf
2019-07-13T08:28:07.811691263Z   File "/srv/conda/envs/notebook/lib/python3.6/site-packages/gin/tf/__init__.py", line 68, in <module>
2019-07-13T08:28:07.811750133Z     from gin.tf.utils import GinConfigSaverHook
2019-07-13T08:28:07.811756881Z   File "/srv/conda/envs/notebook/lib/python3.6/site-packages/gin/tf/utils.py", line 34, in <module>
2019-07-13T08:28:07.811855121Z     config.register_file_reader(tf.io.gfile.GFile, tf.io.gfile.exists)
2019-07-13T08:28:07.811860925Z AttributeError: module 'tensorflow._api.v1.io' has no attribute 'gfile'

https://gitlab.aicrowd.com/STAR.Lab/obstacle-tower-challenge/issues/54

We try to reset our repo to the last successful upload model and codes, but the error remain.
We try to fix this error according to https://github.com/google/gin-config/issues/9 by changing gin-config version to 0.1.4, this error remains still.

Anyone has idea how to fix this ?

#2

We tried a few settings in setup.py, our last attempt success in uploading in debug mode using tensorflow-gpu==1.14

Not sure why since we trained in tensorflow-gpu==1.12