add precision arguments explicitly to examples

[njzjz]

Set precision=float32 for NNs in examples, since we have found that float64 and float32 have the same accuracy in almost all the situation.

[codecov]

Codecov Report

Patch coverage has no change and project coverage change: -0.04 ⚠️

Comparison is base (37fd8d1) 78.35% compared to head (c7cbe03) 78.32%.

Additional details and impacted files

@@            Coverage Diff             @@
##            devel    #2659      +/-   ##
==========================================
- Coverage   78.35%   78.32%   -0.04%     
==========================================
  Files         235      235              
  Lines       24473    24473              
  Branches     1469     1469              
==========================================
- Hits        19177    19169       -8     
- Misses       4906     4914       +8     
  Partials      390      390              

see 7 files with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[wanghan-iapcm]

I would recommend all examples preserves the old behavior and provide separate examples for float32.
The same training/validating accuracy can not justify the same accuracy in real applications, for example, the calculations of elastic constant requires the high accuracy of second order derivatives of the energy. We have relatively little knowledge on the numerical influence on these properties.

[njzjz]

I would recommend all examples preserves the old behavior and provide separate examples for float32.

I change float32 to float64. People can easily change float64 to float32 after we explicitly show the location of these arguments.