Can not run DPA-2 pretrained model for finetuning

[Zch102xjtumse]

I want to use the pretrained model for finetuning basing my own metals datasets in 'https://www.aissquare.com/models/detail?pageType=models&name=DPA-2.2.0-v3.0.0b3&id=272' and the JSON file was obtained in 'https://bohrium.dp.tech/notebooks/2649266844' which I replaced the training sets and validation sets with my own dataset.But it doesn't work.The error is 'dargs.dargs.ArgumentValueError: [at location model/descriptor/list/1] get invalid choice se_uni for flag key type.Did you mean: se_t?'
Is there any wrong with my input JSON file?

[njzjz]

This JSON file is incompatible with DeePMD-kit.

@Chengqian-Zhang could you make clarification in your notebook?

[Chengqian-Zhang]

It's a version issue, I'll update the notebook version today

[Chengqian-Zhang]

Hi, you can read this notebook: https://bohrium.dp.tech/notebooks/18475433825. The version of deepmd-kit for this notebook is beta3.

[Zch102xjtumse]

Hi，I'm grateful for your assistance,but when using the input.json for singletask, I met this problem: dargs.dargs.ArgumentKeyError: [at location model/descriptor] undefined key tebd_dim is not allowed in strict mode. What should I do to correct it?

[Chengqian-Zhang]

I doubt that the deepmd-kit version and input.json version are not consistent. You can check your deepmd-kit version and try using the corresponding examples/water/dpa2/input_torch.json.