Open Soudeh-Jahanshahi opened 5 months ago
I want to add to this report my second concern:
create_document_embeddings
of script embeddings.py
: Last week when I executed the compact-code on de-NBI cloud I could get the results of all 18 sets at once, but using this new modified code I just had space to get the results of first 6 sets at once ...And I really enjoyed using script show_avg.py
, which produces a summary table of average results of all sets for each evaluation quantity --the same as the tables that we should fill in the spreadsheets-- and save it as a tsv file: So far I did fill all my tables in the spreadsheets row by row :)
Everything works fine.
My only concern is about the order of hyperparameter sets which is different from hyperparameters_doc2vec.json, which is the same JSON file that I used for Word2Vec models where sg was replaced by dm.
For instance, the first 9 hyperparameter sets are for
dm=0
, while in hyperparameters_doc2vec.json, the even sets are fordm=0
.