0.2.0
Full Changelog: 0.1.1...0.2.0
- Improves sentence generation performance by loading the text model into a cached property.
- Cached properties are cleared when calling
refresh_from_db
.
- Cached properties are cleared when calling
- Enables storing non-compiled models to the database for use in combining multiple models into a single chain.
- Adds boolean settings parameter
MARKOV_STORE_COMPILED_MODELS
to set default behavior. - This can be overridden when calling
update_model_from_corpus
using thestore_compiled
kwarg. - Adds ability to auto-detect if a loaded model is compiled or not, and will automatically compile for sentence generation if needed.
- Adds boolean settings parameter
- Adds support for the
tries
directive when callinggenerate_sentence
. - Adds support for combining multiple models into one via
MarkovTextModel.combine_models