imread fix (catalyst-team#514)

* Update image.py

* Update image.py

* Update image.py

* Update image.py

* Update image.py

* Update test_image.py

* Update test_image.py

* Update test_image.py

* Update test_image.py

* Update .travis.yml

* fix

* pytest fix

Chore/default callbacks for ConfigExperiment (catalyst-team#505)

* Test for BaseExperiment defaults

* Test for SupervisedExperiment defaults

* Add default callbacks for ConfigExperiment and tests

* Make criterion and optimizer callbacks defined by default in non-infer
mode for SupervisedExperiment

* Fix style

* Some docstrings

* Fix default callbacks definition for case with phase wrapper callbacks

* Fix docs styles

Revet model.to(device) call before amp.initialize (catalyst-team#503)

fix extra / (catalyst-team#484)

* fix extra /
* wrap into Path

add MaskReader (catalyst-team#446)

* add MaskReader

* fix MaskReader import

* fix codestyle

added segmentation tutorial (catalyst-team#415)

* added segmentation tutorial

* TTA

* correct rename

* notebbok

* Update __version__.py

Andreyzharkov multimodel multiphase gan example (catalyst-team#407)

* gan example - somehow working shitcode

* refactoring p1 - maybe_recursive_call

* refactoring p2 - more OOP-like experiment & runner in example

* visualization

* creepy GAN mnist

* codestyle compatibility

* codestyle v2

* multiple models & optimizers by _key_value in ConfigExperiment

* PhaseManagerCallback + remove redundant experiments & runners

* callback wrappers for enabling/disabling handlers dependant on state.phase

* readme

* codestyle again

* make codestyle

* unpack checkpoint fix

* proper load multiple optimizers from checkpoint

* review fix1

* make codestyle

* review fix2

* codestyle

* refactoring

* config.py

* config update

* config update

* refactoring

* reproducible gan

* trust me, i am engineer

* kv+gpu fix

* Update __version__.py

Add checkpoints 16+mb support to MongoDB (catalyst-team#396)

* small fix

* small fix

* Add checkpoints 16+mb support  to MongoDB

* Update mongo.py

* Update __version__.py

Supervised runner resume feature (catalyst-team#377)

* Update supervised.py

* Update supervised.py

* Update supervised.py

Feature. Criterion Aggregator (catalyst-team#361)

* to get a full dictionary if the key is None

* different strategies to aggregate multiple losses

* save loss_aggregate_fn name

* added CriterionAggregatorCallback

* fixes

* fix one-key in dict

* update _get

* loss_key to OptimizerCallback