Add base model for multi-table statistic model, change single-table base class location

[MooooCat]

Description

• Add base model for multi-table statistic model.
• Change single-table base class location.
• Change check settings in metadata.

Motivation and Context

The main purpose of this change is to adapt to multi-table synthetic data scenarios

How has this been tested?

A test case has been provided for the method of base class.

Types of changes

• Maintenance (no change in code, maintain the project's CI, docs, etc.)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)

Checklist:

• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.

[codecov-commenter]

Codecov Report

Attention: 17 lines in your changes are missing coverage. Please review.

Comparison is base (f18b552) 78.68% compared to head (01b1749) 78.99%.

Files	Patch %	Lines
sdgx/models/statistics/multi_tables/base.py	83.58%	11 Missing ⚠️
sdgx/models/base.py	76.92%	3 Missing ⚠️
sdgx/data_models/metadata.py	50.00%	2 Missing ⚠️
sdgx/models/statistics/single_table/base.py	0.00%	1 Missing ⚠️

❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #102      +/-   ##
==========================================
+ Coverage   78.68%   78.99%   +0.30%     
==========================================
  Files          61       62       +1     
  Lines        2599     2680      +81     
==========================================
+ Hits         2045     2117      +72     
- Misses        554      563       +9     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[Wh1isper]

Nice PR, and I have a couple questions:

• Why does MultiTableSynthesizerModel inherit from pydantic.BaseModel, given that he doesn't seem to be a data object.
• If you want to accept streaming files(DataLoader) or Dataframe for evaluation, you only need to define the two mutually exclusive parameters raw_data and dataloader in the kwargs, what is the importance of the enum class DataAccessType?
  • Also, if the raw_data parameter needs to be added, one might want to consider how the existing architecture extends this parameter

[MooooCat]

• Why does MultiTableSynthesizerModel inherit from pydantic.BaseModel, given that he doesn't seem to be a data object.

Regarding this question, it mainly comes from my lack of understanding of pydantic.BaseModel. Before that, I thought it was just a base class convenient for developer to implement. After checking, I found that:

"Pydantic is the most widely used data validation library for Python. ”

this is indeed inconsistent with the purpose of a statistical model.

[Z712023]

Offer fuel on snowy weather! Multi-table model is what we want!

[Z712023]

The validation in metadata is ok!

[MooooCat]

Offer fuel on snowy weather! Multi-table model is what we want!

There will be some minor modifications before merge.

[MooooCat]

• Why does MultiTableSynthesizerModel inherit from pydantic.BaseModel, given that he doesn't seem to be a data object.
• If you want to accept streaming files(DataLoader) or Dataframe for evaluation, you only need to define the two mutually exclusive parameters raw_data and dataloader in the kwargs, what is the importance of the enum class DataAccessType?
  • Also, if the raw_data parameter needs to be added, one might want to consider how the existing architecture extends this parameter

I removed the pydantic.BaseModel, and use sdgx.models.base.SynthesizerModel instead.

I removed the DataAccessType and set two mutually exclusive parameters use_raw_data and use_dataloader in sdgx.models.base.SynthesizerModel, this base class will be inherit by all ML and statistic models.

In existing architecture, I think it's OK to simply set use_raw_data to True or use_dataloader to True.

In the future, we can add support for these two data access methods . (I leave this to future, not in this PR).

Currently, this modification does not affect the use of existing models.

[Wh1isper]

I prefer to merge it now and make further improvements in the implementation of multi-table simulations.