feat: apply descriptor exclude_types to env mat stat

[njzjz]

No description provided.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 79.10%. Comparing base (4c546d0) to head (a5e9377).

Additional details and impacted files

@@            Coverage Diff             @@
##            devel    #3625      +/-   ##
==========================================
- Coverage   81.71%   79.10%   -2.61%     
==========================================
  Files         503      503              
  Lines       45992    45934      -58     
  Branches     2891     2891              
==========================================
- Hits        37581    36337    -1244     
- Misses       7533     8858    +1325     
+ Partials      878      739     -139     

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

[wanghan-iapcm]

I agree with this PR, but please not that would deprecate the exclude type in the future.

[njzjz]

would deprecate the exclude type in the future.

What's the reason for that?

[wanghan-iapcm]

would deprecate the exclude type in the future.

What's the reason for that?

because not all atomic model has descriptor. the dp atomic model's exclude pair is equivalent to the descriptors exclude pair.

[njzjz]

would deprecate the exclude type in the future.

What's the reason for that?

because not all atomic model has descriptor. the dp atomic model's exclude pair is equivalent to the descriptors exclude pair.

An atomic model may not contain only one descriptor. Note that two descriptors in a DPRc model have different exclude pairs. So it may not be equivalent.

[wanghan-iapcm]

would deprecate the exclude type in the future.

What's the reason for that?

because not all atomic model has descriptor. the dp atomic model's exclude pair is equivalent to the descriptors exclude pair.

An atomic model may not contain only one descriptor. Note that two descriptors in a DPRc model have different exclude pairs. So it may not be equivalent.

I see in multiple descriptor scenario they are not equivalent.

[iProzd]

I notice that the excluded_types only masks the input env_mat, while still keeps the total number of atoms in excluded_types (sumn). Another way is to mask excluded_types in both numerator (sumr etc.) and denominator (sumn) . Which one should we prefer? @wanghan-iapcm @njzjz

[wanghan-iapcm]

I notice that the excluded_types only masks the input env_mat, while still keeps the total number of atoms in excluded_types (sumn). Another way is to mask excluded_types in both numerator (sumr etc.) and denominator (sumn) . Which one should we prefer? @wanghan-iapcm @njzjz

I would prefer mask both the numerator and the denominator, which gives consistent result as if the masked pairs do not exists.

[njzjz]

I notice that the excluded_types only masks the input env_mat, while still keeps the total number of atoms in excluded_types (sumn). Another way is to mask excluded_types in both numerator (sumr etc.) and denominator (sumn) . Which one should we prefer? @wanghan-iapcm @njzjz

I would prefer mask both the numerator and the denominator, which gives consistent result as if the masked pairs do not exists.

They are not equivalent. When masked pairs do not exist, zeros (meaning no neighbor) will be used for the stat.

The current masking behavior has some problems. The env mat is always masked to zero, but if set_davg is false, it should be masked to davg, otherwise, there is discontinuity.