Implemented the parameter chunk_size in the glossary generator, that is also logged to wandb. That way, we get a more coherent parameter compared to “the 100/nth percent of the paper per chunk”, and make it easier to chunk large documents. Also logged the number of generated glossary entries, where we nicely see that with a larger chunk size, the number of entries decreases.
Implemented the parameter
chunk_size
in the glossary generator, that is also logged to wandb. That way, we get a more coherent parameter compared to “the 100/nth percent of the paper per chunk”, and make it easier to chunk large documents. Also logged the number of generated glossary entries, where we nicely see that with a larger chunk size, the number of entries decreases.