- update to 0.4.0:
* Bump model version and fix typo
* Add HashingAnnoy model
* Add hashing\_nn benchmark in doc string
* Add HashingApproximateNeighbors model
* Implement iterator interface for file-like objects
* Refactor TokenizerTests
* Provide a bit more info about timings of the training
* Remove support for bag-of-words\_lshf
* Don't store duplicate data in model
* Fix heat\_uuid regexp formatting
* Relax digits\_re again a bit
* Vectorizer optimisation: don't do word analysing
* debug\_lineprocess: Handle more than one input file
* debug\_lineprocess: Format output slightly nicer and remove duplicates
* Tighten heat\_uuid regexp
* Tighten length-based regexp matches properly
* debug\_lineprocess add some simple word / token statistics
* Blacklist .xml extension
* Use for loop instead of handcrafted while construct
* tests: use free tcp port for gearman server
* Add --model-type argument to top-level command
* tokenizer: remove sshd warnings
* Make debugging scripts callable again
* Reduce code duplication a bit
* Micro-optimize the tokenization
* ci: enable gate jobs
* Make systemd service file SCL independent
* Transition webui related files to the log-classify name
* Match uuid\_re before heat\_re