Oleksandr Frei

Loading a model saved in v0.9.0 fails with transaction artm: ``` import artm import os model=artm.ARTM(num_topics=1) model.load(mymodel, model_name="pwt") ``` Error: ``` --------------------------------------------------------------------------- UnboundLocalError Traceback (most recent call last) in ()...

Workaround weird build issue on manylinux docker image (centos 5)

1

The ``-lrt`` flag must be the last in linker command, that's why I'm adding it into target_link_libary. I've tried to add into CMAKE_EXE_LINKER_FLAGS, but then it is prepended somewhere in...

BatchVectorizer fails to parse VW format (unless target folder is specified)

1

``` import artm batch_vectorizer = artm.BatchVectorizer(data_path=r'H:\datasets\vw.eurlex.txt', data_format='vowpal_wabbit') ``` => ``` --------------------------------------------------------------------------- InternalError Traceback (most recent call last) in () 1 import artm ----> 2 batch_vectorizer = artm.BatchVectorizer(data_path=r'H:\datasets\vw.eurlex.txt', data_format='vowpal_wabbit') C:\Users\Oleksandr\Documents\GitHub\bigartm\python\artm\batches_utils.py in...

Функция dictionary.gather(data_path, cooc_file_path, vocab_file_path, symmetric_cooc_values) не падает и не выдает никаких предупреждений, если в качестве vocab_file_path подать путь до несуществующего файла

Implement unbiased regularization

2

As described by I. Irihin in his presentation.

Since recently `RegularizerConfig` message is part of `MasterModelConfig` message, thus reconfiguring regularizers happen automatically from `ArtmReconfigureMasterModel`. However we still an keep individual call to create/dispose/reconfigure regularizer. These call should update...

Recently we've implemented `hARTM` class in python to build hierarchical topic models. We need to push this feature down into BigARTM Core, and make it available in BigARTM CLI. Please...

Online алгоритм, или Ещё более большие данные

8

Тенденция такова, что нам придётся работать на всё больших данных и на динамически пополняемых коллекциях. Все случаи, когда нам дали готовую коллекцию -- это история про научное исследование, пилотный проект...