Test infrastructure for SageMaker (mlflow#718)

* Add SparkJava plugin

* Add MLeap flavor and modify SparkML module to use it

* Fixes

* Add mleap

* Import model in test

* Import mleap

* Add missing assert

* Reorder spark session creation params in test

* Docs fix

* revert pom xml change

* remove java

* Add standalone  to mleap

* Import fix

* Add docs and

* Add warning about py incompatibility

* Address comments

* Code spacing fix

* Revert log test changes

* Whitespace fixes

* Exception import fixes

* Fix lint issues

* Fix method call

* callfix

* Remove unused imports

* Py4j log level fix

* reorder tests

* testfix

* test fixes

* Spacing fix

* lint fixes

* Add mleap schema test

* Fix test

* Whitespace fix

* Test fix

* Bump version to 0.5.3dev0

* Update container + build function to support MLeap

* Fix lint errors

* Install maven

* Test support infra

* Fixws

* Add tests

* Path fixes

* line fix

* Lint fixes

* Revert scoring server

* Update test asserts, remove unused file

* Specify server port

* Add mock for sagemaker functions that we use

* Sigquit >> sigterm

* accept any version of mlflow jar

* Address some comments, fix bug in scoring server argument parsing (whoops :/), add support for specifying deployment flavor

* Address comments

* Update ScoringServer.java

* Update ScoringServer.java

* Update __init__.py

* varfix

* Revert tempdir changes

* Mark test as large

* Use sagemaker env var instead of file upload

* Parameter fixes

* CLI fixes

* CLI Fixes

* CLI support, tests

* Make linter happy

* Fix serving command

* Name change for get_serving_flavor

* Address comments

* Make linter happy

* Address comments

* Add  cli command and refactor flavor selection/validation

* Test fixes

* Add print statements

* Varfix

* Run local in shell to get docker env var successfully

* Lint fix

* Iteration fix

* Deployment env var fixes for local serving

* Sagemaker mock progress

* Rename service mock

* Mock test

* Lint

* Revert extraneous files

* Revert

* Tests

* Mock tests and fixes

* Better test coverage

* Update test

* Docs

* Revert examples

* Use mocks in real tests

* Lint

* lint

* Add mocks test

* Lint

* Unpin moto and boto

* Fix doc spacing

* Add mock boto credentials

* Fix tests

* Lint

* Test fix

* Lint

* Check more specific exception

* Wraps fix

* Lint fix

* Remove needless encoding

.travis.yml

@@ -99,6 +99,7 @@ script:
   - pytest --verbose tests/pytorch --large
   - pytest --verbose tests/pyfunc --large
   - pytest --verbose tests/sagemaker --large
+  - pytest --verbose tests/sagemaker/mock --large
   - pytest --verbose tests/sklearn --large
   - pytest --verbose tests/spark --large
   - pytest --verbose tests/tensorflow --large

test-requirements.txt

@@ -2,10 +2,9 @@
 azure-storage
 google-cloud-storage
 h2o
-# TODO: don't pin boto version once https://github.com/spulec/moto/issues/1793 is addressed
-boto3==1.7.84
+boto3
 mock==2.0.0
-moto==1.3.4
+moto
 prospector[with_pyroma]==0.12.7
 pep8==1.7.1
 pyarrow