open-hyped / hyped / 10351415832

97%

Build Type

push

github

Committed by web-flow

Commit Message

Refactor/data flow (#52)

* Implemented graph based data flow

* Implement tests for feature ref and related components

* Rework data flow and processor syntax

* Implement tests for data flow and related components

* Implement tests for base data processor

* Implement tests for noop and collect features processors

* Run Formatter and Linter

* Fill in docstrings

* Introduce flow module

* Rework node and edge properties

* Add pytest-asycio to test dependency

* Implement plot function for data flow objects

* Implement json string parser processor

* Clean up error message for feature type mismatch

* Restructure package in preparation for augmentors

* Rework docstrings

* Added missing import

* Ported the openai chat completion processor

* Update README.md

* Update README.md

* Fix bugs in FeatureCollector processor

* Edit type annotations in collect.

* Fix linting.

* Refactor feature collector

* Implement constant feature support in feature collection

* Update docstrings

* Add tests and support manually specifying the feature type of constant features

* Use path in nested collection as name instead of hash

* Support optional input features

* Support testing for output feature type

* Implement transformers tokenizer processor

* Implement lazy instance that recreates it's instance for every new process or event loop

* Make semaphore lazy

* Support initializing processor without config

* Catch error on patch asyncio

* Type hints for configs, input and output refs; Fix filter warnings in lazy instance with python<3.12

* Refactor Jinja2 to work with DataFlow (#36)

* Jinja2 flow processor.

* Unit tests for Jinja2 processor.

* Switch jinja implementation to LazyInstances.

---------

Co-authored-by: Henri Iser <henri.iser@iais.fraunhofer.de>

* Feature/json parser error handling (#49)

* Fix base data processor batch_process expects condition... (continued)

Coverage Stats

2161 of 2187 new or added lines in 38 files covered. (98.81%)

8 existing lines in 4 files now uncovered.

2903 of 2980 relevant lines covered (97.42%)

0.97 hits per line