Fix common issues #68

MorrisNein · 2023-11-01T00:08:11Z

Solves following issues:

pep8speaks · 2023-11-01T00:08:31Z

Hello @MorrisNein! Thanks for updating this PR. We checked the lines you've touched for PEP 8 issues, and found:

In the file meta_automl/meta_algorithm/dataset_similarity_assessors/__init__.py:

Line 1:1: F401 '.dataset_similarity_assessor.DatasetSimilarityAssessor' imported but unused
Line 2:1: F401 '.model_based_similarity_assessors.KNeighborsSimilarityAssessor' imported but unused
Line 2:1: F401 '.model_based_similarity_assessors.ModelBasedSimilarityAssessor' imported but unused

In the file meta_automl/meta_algorithm/model_advisors/__init__.py:

Line 1:1: F401 'meta_automl.meta_algorithm.model_advisors.model_advisor.DatasetSimilarityModelAdvisor' imported but unused
Line 1:1: F401 'meta_automl.meta_algorithm.model_advisors.model_advisor.ModelAdvisor' imported but unused
Line 2:1: F401 'meta_automl.meta_algorithm.model_advisors.diverse_model_advisor.DiverseModelAdvisor' imported but unused
Line 3:1: F401 'meta_automl.meta_algorithm.model_advisors.surrogate_advisor.SurrogateGNNPipelineAdvisor' imported but unused

Comment last updated at 2023-11-07 14:41:19 UTC

codecov · 2023-11-01T00:14:03Z

Codecov Report

Merging #68 (5edec59) into main (069719f) will increase coverage by 0.25%.
The diff coverage is 26.97%.

❗ Current head 5edec59 differs from pull request most recent head 4061730. Consider uploading reports for the commit 4061730 to get more accurate results

@@            Coverage Diff             @@
##             main      #68      +/-   ##
==========================================
+ Coverage   28.50%   28.76%   +0.25%     
==========================================
  Files          53       53              
  Lines        2319     2347      +28     
==========================================
+ Hits          661      675      +14     
- Misses       1658     1672      +14

Files	Coverage Δ
...ta_preparation/datasets_loaders/datasets_loader.py	`88.88% <100.00%> (ø)`
...tion/datasets_loaders/timeseries_dataset_loader.py	`100.00% <100.00%> (ø)`
...ion/feature_preprocessors/feature_preprocessors.py	`25.00% <100.00%> (+1.74%)`	⬆️
...automl/data_preparation/file_system/file_system.py	`90.00% <100.00%> (+1.76%)`	⬆️
...ration/meta_features_extractors/pymfe_extractor.py	`67.81% <100.00%> (-0.37%)`	⬇️
meta_automl/data_preparation/evaluated_model.py	`0.00% <0.00%> (ø)`
...ion/models_loaders/knowledge_base_models_loader.py	`0.00% <0.00%> (ø)`
...algorithm/dataset_similarity_assessors/__init__.py	`0.00% <0.00%> (ø)`
...imilarity_assessors/dataset_similarity_assessor.py	`0.00% <0.00%> (ø)`
...a_automl/meta_algorithm/model_advisors/__init__.py	`0.00% <0.00%> (ø)`
... and 9 more

ShikovEgor

В /experiments/base хранились чекпоинты обученной модели для использования в дальнейшем. Название директорий так себе, но на мой взгляд в /data им тоже не место.
Предлагаю создать директорию model_checkpoints (либо model_weights) в корне. Примерная структура:
model_checkpoinst
______ table_data
____________ checkpoints
____________ events.out.tfevents…..
____________ hparams.yaml
______ timeseries
.....

valer1435 · 2023-11-09T07:33:29Z

examples/6_inference/inference_ts_knn.py

-    x_train, x_test = train_test_split(meta_features, train_size=0.75, random_state=42)
-    y_train = x_train.index
-    y_test = x_test.index
+    mf_train, mf_test = train_test_split(meta_features, train_size=0.75, random_state=42)


Здесь тогда тоже можно разделить вместе с индексом сразу

Сделал

valer1435 · 2023-11-09T07:36:31Z

meta_automl/data_preparation/meta_features_extractors/pymfe_extractor.py

+        for idx, col in enumerate(x.columns):
+            is_categorical = cat_cols_indicator[idx]
+            if is_categorical:
+                most_frequent = x_new[col].value_counts(sort=True, ascending=False).values[0]


Мб информация о том, что не удалось посчитать данный признак более информативна, чем просто заменить все на моду и медиану?

Т.е. исключать целиком признак датасета из расчёта мета-признаков, если признак встречает none?

Пока что заменил на более лаконичный расчёт моды

Нет. Я имел ввиду, что мы можем заменить нан на какое-то специальное значение, которое будет говорить о том, что значения там нет.

Но такое значение, чтобы модель смогла переварить

valer1435 · 2023-11-09T07:38:25Z

...data_preparation/meta_features_extractors/time_series/time_series_meta_features_extractor.py

+                input_data = InputData(idx=np.array([0]), features=np.array(features).reshape(1, -1), target=None,
+                                       task=Task(TaskTypesEnum.classification),
+                                       data_type=DataTypesEnum.table)
+                with IndustrialModels():


Это лучше перенести вне цикла, чтобы не вызывать каждый раз тяжелую операцию подгрузки репозитория

Попробовал убрать, всё работает. Контекст оказался нужен только при загрузке пайплайна

valer1435 · 2023-11-10T12:52:35Z

meta_automl/data_preparation/meta_features_extractors/pymfe_extractor.py

            else:
-                median = x_new[col].median()
-                x_new[col].fillna(median, inplace=True)
+                fill_value = x_new[col].median(skipna=True)


Можно оставить и как было. Про замену нанов каким-то специфичным значением просто идея

MorrisNein added 8 commits October 31, 2023 22:20

fix inner components

dceeaab

fix time series components

205e84f

separate interface of DatasetSimilarityAssessor and ModelAdvisor

76a74e1

fix examples

d15eb98

fix typos

3a39bc8

fix feature extractors

3d7808d

make abstract classes inherit ABC

e54f034

rename Model to EvaluatedModel

3f4f656

MorrisNein requested review from valer1435 and ShikovEgor November 1, 2023 00:08

MorrisNein force-pushed the fix-common-issues branch from 18b285a to c4d8ec7 Compare November 1, 2023 00:10

ShikovEgor reviewed Nov 1, 2023

View reviewed changes

ShikovEgor approved these changes Nov 2, 2023

View reviewed changes

MorrisNein added 7 commits November 2, 2023 21:55

fix path to surrogate knowledge base

1ff7227

fix test_file_system.py, add test_cache.py

7dd7dc1

use Path instead of str

dcc9e58

pep8

a2b40fe

pep8 & minor fixes

85bb439

fix classes inheritance

e2e911f

fix ts example

222d693

MorrisNein force-pushed the fix-common-issues branch from 5edec59 to 222d693 Compare November 2, 2023 18:57

MorrisNein added 6 commits November 2, 2023 22:13

add get_checkpoints_dir(), fix examples

5a8ca7c

add test_checkpoints_dir

28cd9a2

pep8

be39141

delete inconsistent example

b26d8e5

fix type hints

95e9d7b

fix logging

bb3f03a

MorrisNein force-pushed the fix-common-issues branch from 39f9557 to bb3f03a Compare November 7, 2023 14:40

valer1435 requested changes Nov 9, 2023

View reviewed changes

MorrisNein mentioned this pull request Nov 9, 2023

Move to fedot 0.7.3 and golem 0.3.3 #79

Closed

MorrisNein added 7 commits November 9, 2023 22:03

use __init__.py files for time series components

56c56e4

update the archive

098d594

fix arbitrary path at advise_by_surrogate.py example

a8a1772

split index with df

f8d7f22

remove unnecessary context

0fdbbc9

better mode computation

6a20d75

test input nans filling by default

4061730

MorrisNein requested a review from valer1435 November 9, 2023 20:49

valer1435 reviewed Nov 10, 2023

View reviewed changes

valer1435 approved these changes Nov 10, 2023

View reviewed changes

MorrisNein force-pushed the fix-common-issues branch from dc6c069 to 4061730 Compare November 10, 2023 13:49

MorrisNein merged commit 150d53c into main Nov 10, 2023

MorrisNein deleted the fix-common-issues branch November 10, 2023 13:51

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Fix common issues #68

Fix common issues #68

MorrisNein commented Nov 1, 2023 •

edited

Loading

pep8speaks commented Nov 1, 2023 •

edited

Loading

codecov bot commented Nov 1, 2023 •

edited

Loading

ShikovEgor left a comment •

edited

Loading

valer1435 Nov 9, 2023

MorrisNein Nov 9, 2023

valer1435 Nov 9, 2023

MorrisNein Nov 9, 2023

MorrisNein Nov 9, 2023

valer1435 Nov 10, 2023

valer1435 Nov 9, 2023

MorrisNein Nov 9, 2023

valer1435 Nov 10, 2023

Fix common issues #68

Fix common issues #68

Conversation

MorrisNein commented Nov 1, 2023 • edited Loading

pep8speaks commented Nov 1, 2023 • edited Loading

Comment last updated at 2023-11-07 14:41:19 UTC

codecov bot commented Nov 1, 2023 • edited Loading

Codecov Report

ShikovEgor left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

MorrisNein commented Nov 1, 2023 •

edited

Loading

pep8speaks commented Nov 1, 2023 •

edited

Loading

codecov bot commented Nov 1, 2023 •

edited

Loading

ShikovEgor left a comment •

edited

Loading