Simplify deploy and rename .yml -> .yaml for consistency.

In models.py update the on_delete constraint in region
for frame deletion to null. Further simplify deployment.

Soon:

- YFCC processing and preparation scripts to VDN.

- Export trained approximators as zip file.

Next:

- Implement pre-condition checks, failed to launch error
messages for processes. Rollback created models on failure to launch.

- Add location net, fine-tuning / training face recognition models

deploy/compose/__init__.py

@@ -109,7 +109,7 @@ def stop_docker_compose(deployment_type, gpu_count, clean=False):
     try:
         subprocess.check_call(["docker-compose", '-f', fname, 'down'] + extra_args,
                               cwd=os.path.join(os.path.dirname(os.path.curdir),
-                                               'deploy/compose/{}'.format(deployment_type)))
+                                               'deploy/compose'))
     except:
         raise SystemError("Could not stop containers")