Empty: Failed to retrive number of trained models

[machine-intelligence-laboratory]

Empty: Failed to retrive number of trained models

[Evgeny-Egorov-Projects]

Hi there!

Seems like in your case our multithreading approach failed to finish training a model before the program tried to extract it to the parent process. A quick way to fix this issue would be to use
separate_thread=False option in:

my_first_cube = RegularizersModifierCube(
    num_iter=5,
    tracked_score_function='PerplexityScore@lemmatized',
    regularizer_parameters={
        'regularizer': artm.DecorrelatorPhiRegularizer(name='decorrelation_phi', tau=1),
        'tau_grid': [0,1,2,3,4,5],
    },
    reg_search='grid',
    separate_thread=False,
    verbose=True
)

Could you, please tell a bit more about the dataset you used and the system configuration you trained it on?

[kamil-safin]

Hi
I use dataset of chat conversation with about 6k short dialogues. My system is MasOS with topicnet==0.4.1. I'm following your guide but use my own dataset.
And now I have:
InvalidOperationException: TopicKernelScoreConfig.class_id @bigram does not exists in n_wt matrix

[Evgeny-Egorov-Projects]

Did you use our preprocessing guide?
Can you confirm that you have modality named '@bigram' in your dataset?
If you have that modality can you check if it is not empty?

[kamil-safin]

Yes, I used
Example of single line in dataset ('vw_text' field):
6 |@lemmatized подать:1 новый:1 заявка:1 |@bigrams подать_новый:1 новый_заявка:1
How can check the existence of modality in my model?

[kamil-safin]

sorry, my bad:
used '@bigram' and '@bigrams' in different places of code
thanks for separate_thread solution!