get "undefined key "stop_batch" is not allowed in strict mode" error

[taipinghu]

Summary

the above error occurs when init-model is used.
after changing "numb_steps" to "stop_batch", it works normally.

DeePMD-kit Version

2.1.5

TensorFlow Version

default version in deepmd kit 2.1.5

Python Version, CUDA Version, GCC Version, LAMMPS Version, etc

bohrium contain name: registry.dp.tech/dptech/deepmd-kit:2.1.5-cuda11.6

Details

the above error occurs when init-model is used.
after change the keyword numb_steps to stop_batch, it works normally.

[njzjz]

Did you use an old version? stop_batch was renamed to numb_steps in deepmodeling/deepmd-kit#502.

[taipinghu]

Did you use an old version? stop_batch was renamed to numb_steps in deepmodeling/deepmd-kit#502.

the version is 2.1.5.
numb_steps works normally when init-model is not used.

[wanghan-iapcm]

Did you train the model, from which you initialize your new model, by the latest version of deepmd-kit?

[taipinghu]

Did you train the model, from which you initialize your new model, by the latest version of deepmd-kit?

Firstly, I make a long train in iter 0. Here, the numb_steps works normally.
Then, in iter 1, I use init model based on the previous model. Here, deepmd kit automatically convert input.json to input_v2_compat.json. it cannot find "stop_batch" keyword in the input_v2_compat.json. I change the param.json (dpgen param) from the numb_steps to stop_batch, "stop_batch" appears in the input_v2_compat.json and deepmd kit works normally.

[njzjz]

This should be a bug of DP-GEN. I will transfer the issue there.

[AnguseZhang]

This seems solved. I'll close this issue for now.