Dependencies not downloading on GCP

[alexf-a]

Thanks for the tf.keras support!

I am trying to implement segmentation_models with tf.keras on GCP, so am trying to reference the latest version in my GCP AI-Platform setup.py package with "segmentation_models>=1.0.0b1".

When I call set_framework to use tf.keras, I get:

File "/usr/lib/python3.5/runpy.py", line 184, in _run_module_as_main
ERROR	2019-09-26 12:57:02 -0400	service		    "__main__", mod_spec)
ERROR	2019-09-26 12:57:02 -0400	service		  File "/usr/lib/python3.5/runpy.py", line 85, in _run_code
ERROR	2019-09-26 12:57:02 -0400	service		    exec(code, run_globals)
ERROR	2019-09-26 12:57:02 -0400	service		  File "/root/.local/lib/python3.5/site-packages/trainer/task.py", line 9, in <module>
ERROR	2019-09-26 12:57:02 -0400	service		    from .task_utils import parse_args, get_args
ERROR	2019-09-26 12:57:02 -0400	service		  File "/root/.local/lib/python3.5/site-packages/trainer/task_utils.py", line 9, in <module>
ERROR	2019-09-26 12:57:02 -0400	service		    import segmentation_models
ERROR	2019-09-26 12:57:02 -0400	service		  File "/root/.local/lib/python3.5/site-packages/segmentation_models/__init__.py", line 101, in <module>
ERROR	2019-09-26 12:57:02 -0400	service		    set_framework(other)
ERROR	2019-09-26 12:57:02 -0400	service		  File "/root/.local/lib/python3.5/site-packages/segmentation_models/__init__.py", line 71, in set_framework
ERROR	2019-09-26 12:57:02 -0400	service		    import efficientnet.tfkeras  # init custom objects
ERROR	2019-09-26 12:57:02 -0400	service		ImportError: No module named 'efficientnet'
ERROR	2019-09-26 12:57:02 -0400	service	

[qubvel]

Hi this is not GCP problem) Something wrong with travis deploy, I will fix it in next version, thanks for report!