onnx / turnkeyml

The no-code AI toolchain
Apache License 2.0
63 stars 12 forks source link

Refresh: Eliminate the state dataclass and state.yaml file #152

Closed jeremyfowers closed 3 months ago

jeremyfowers commented 3 months ago

The State class and state.yaml have slowed down development and caused compatibility issues by being overly rigid. In the future, all evaluation statistics will be held in stats.yaml and all evaluate state will be passed between Stages in a state: dict.

jeremyfowers commented 3 months ago

A big part of why state.yaml is important is to support a robust caching system. We may update the scope here to include removing the cache hit/miss system, since we can't think of any good reason to keep maintaining it.

cc @danielholanda

jeremyfowers commented 3 months ago

Closed by #162