Yingze Wang comments

Results 11 comments of


                                            Yingze Wang

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

Hi @Cynthia0807-learn , how did you get the `graph.pb` file? It seems to be something wrong with deepmd parsing the model. I try to load the file directly from deepmd...

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

@Cynthia0807-learn Hi, you mean that deepmd-kit is installed from gromacs-dp offline package provided here?https://github.com/Ericwang6/gromacs-deepmd-feedstock And could you please give a detailed information of your machine? maybe the type of GPU?...

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

I have tried the methane model in the example. And it works fine

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

[frozen_model.pb.zip](https://github.com/deepmodeling/deepmd-kit/files/8225710/frozen_model.pb.zip) This graph should work

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

I see. Plz give me some time to further check this

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

Sure. `type.raw` follows `dpdata.System` conventions and it is defined according to `type_map` in deepmd training json file.

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

Is the problem with `protobuf` solved? I think this is an potential bug in the conda-build process in gromacs-dp

[BUG] An error occured when running gmx-dp:* Error in `gmx': free(): invalid next size (fast): 0x000055b6648a0130 *

> This model is for CH4 and trained by deepmd-kit 2.0.1 which is installed from off-linepackage.

[Parameters] Why are there tow "type_map" items in param.json?

They should be the same and both are required....I'll fix them and make a pr

RuntimeError: Job 634fdaf9-f361-4482-96bd-a525e834a435 failed for more than 3 times

It seems happen during model training, please check `train.log` for specific error infos in your work path `/home/jiangjun/cptwokile/1/c1/data/`