fix 'NoneType' has no len() in auto_sel

[njzjz]

The default value of type_map is None, so when you don't set type_map, you'll get this error.

deepmd-kit/deepmd/entrypoints/train.py

Line 225 in 043ac86

return jdata['model'].get('type_map', None)

[codecov-commenter]

Codecov Report

Merging #911 (0996d05) into devel (31f1ef6) will decrease coverage by 0.07%.
The diff coverage is 0.00%.

@@            Coverage Diff             @@
##            devel     #911      +/-   ##
==========================================
- Coverage   74.65%   74.57%   -0.08%     
==========================================
  Files          85       85              
  Lines        6730     6745      +15     
==========================================
+ Hits         5024     5030       +6     
- Misses       1706     1715       +9     

Impacted Files	Coverage Δ
deepmd/entrypoints/train.py	75.00% <0.00%> (ø)
deepmd/entrypoints/doc.py	33.33% <0.00%> (-26.67%)	⬇️
deepmd/descriptor/se_a.py	97.56% <0.00%> (-1.37%)	⬇️
deepmd/utils/argcheck.py	87.54% <0.00%> (-0.23%)	⬇️
deepmd/descriptor/se_r.py	92.85% <0.00%> (ø)
deepmd/entrypoints/main.py	88.42% <0.00%> (+0.12%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 31f1ef6...0996d05. Read the comment docs.

[njzjz]

By the way, can we let the auto-sel calculation become a default behavior, and throw a RuntimeError when it's more than those set by users? Usually we will get a bad model if sel is not enough for training data. It wastes time to train such a model.

[amcadmus]

By the way, can we let the auto-sel calculation become a default behavior, and throw a RuntimeError when it's more than those set by users? Usually we will get a bad model if sel is not enough for training data. It wastes time to train such a model.

We should allow the users setting a small sel and warn them that they are using a too small sel and they really know what they are doing.