There is no rms information for the test set in the lcurve.out #781
Description
Summary
There is only the rms information for the training set, but no rms information for the test set in the lcurve.out.
We are using the deepmd-kit-2.0.0.b1-cuda11.1_gpu version to train a potential energy surface. The dp train can run normally, and the training process also seems normal. However, in the lcurve.out during the training procedure, there is no output of the rms information for the test set, but only the rms for training set as the following,
" # step rmse_trn rmse_e_trn rmse_f_trn lr
0 2.77e+01 6.40e-01 8.74e-01 1.0e-03
100 1.39e+01 2.22e-01 4.39e-01 1.0e-03
200 8.11e+00 1.61e-01 2.56e-01 1.0e-03
300 6.70e+00 2.64e-02 2.12e-01 1.0e-03
400 8.27e+00 6.61e-03 2.61e-01 1.0e-03
500 5.81e+00 1.18e-02 1.84e-01 1.0e-03"
Normallly in the previous versions, the output in the lcurve.out is like the following,
"# batch l2_tst l2_trn l2_e_tst l2_e_trn l2_f_tst l2_f_trn lr
0 3.84e+01 3.95e+01 1.03e+01 1.03e+01 8.22e-01 8.74e-01 1.0e-03
100 2.68e+01 2.66e+01 6.05e-01 5.98e-01 8.46e-01 8.38e-01 1.0e-03
200 2.51e+01 2.58e+01 8.94e-02 8.85e-02 7.95e-01 8.16e-01 1.0e-03
300 2.16e+01 2.05e+01 7.43e-02 7.17e-02 6.82e-01 6.49e-01 1.0e-03
400 1.73e+01 1.76e+01 2.43e-02 3.02e-02 5.48e-01 5.56e-01 1.0e-03
500 1.49e+01 1.44e+01 1.14e-01 1.08e-01 4.70e-01 4.55e-01 1.0e-03"
So, I am wondering whether there is a bug in deepmd-kit-2.0.0.b1-cuda11.1_gpu-Linux-x86_64, or my input is not correct. My input file has been attached. Please help to check it. Thanks very much !
Deepmd-kit version, installation way, input file, running commands, error log, etc.
deepmd-kit-2.0.0.b1-cuda11.1_gpu-Linux-x86_64
cuda11.1
Steps to Reproduce
Further Information, Files, and Links