Gordo

- Optionally (default=True) use parquet for data transfer between client and server (472)
- Fix issue where models with functions as default parameters did not load properly (477)

- Change sleep time from xor to exponentiation (https://github.com/equinor/gordo-components/pull/465)
- Builder: always save model to output_dir (https://github.com/equinor/gordo-components/pull/461)
- Change retries in client to have exponential sleep time (https://github.com/equinor/gordo-components/pull/464)
- Expose evaluation from config in model builder (https://github.com/equinor/gordo-components/pull/458)
- Make ML Server serve multiple models (https://github.com/equinor/gordo-components/pull/460)

- Update requirements (https://github.com/equinor/gordo-components/pull/440)
- Make autoencoder API parameters match and add some input validation (https://github.com/equinor/gordo-components/pull/431)
- Expose `optimizer` and `optimizer_kwargs` in feedforward_model factory (https://github.com/equinor/gordo-components/pull/429)
- Change interactive logic (https://github.com/equinor/gordo-components/pull/428)
- Cli dockstring fix (https://github.com/equinor/gordo-components/pull/439)
- Add model offset to metadata (https://github.com/equinor/gordo-components/pull/444)
- Model builder - Support a pure CV/HP mode (https://github.com/equinor/gordo-components/pull/395)
- Switch to CircleCI and push to Dockerhub (https://github.com/equinor/gordo-components/pull/442, https://github.com/equinor/gordo-components/pull/451, https://github.com/equinor/gordo-components/pull/450, https://github.com/equinor/gordo-components/pull/454)
- Add EV, R2, MSE and MAE to metadata (https://github.com/equinor/gordo-components/pull/377)
- Client adjust for offset (https://github.com/equinor/gordo-components/pull/445)
- Add more explicit docs about resolution strings (https://github.com/equinor/gordo-components/pull/457)

- Add simple dataframe serialization ability to server serializer utils (https://github.com/equinor/gordo-components/commit/0b2b84dbf93a3e442d78359c3477e4c51974d025)
- Remove logic that y must be a subset of X for anomaly detection. (https://github.com/equinor/gordo-components/commit/6d75c37dc913e89d2bd299d61246be28de684346)

- Make client auto choose prediction endpoint (https://github.com/equinor/gordo-components/commit/66901600183c50a60ae22d65f56189291255384d)
- Add **kwargs to sub_providers (iroc and ncs_reader) (https://github.com/equinor/gordo-components/commit/63c53ef9eade25732f683f757adb9a3ee5ce193b)
- Remove explicit threads=None argument from DataLakeProvider __init__ (https://github.com/equinor/gordo-components/commit/697a5b2d84f6d981eca8d993481f6de113576efe)
- Fix overwrite of y in KerasLSTMAutoEncoder.fit (https://github.com/equinor/gordo-components/commit/dd390e16332d0c84ce0218e9c178aa9e649b50eb)