How to obtain the predicted values of validation set?

[phys-chem]

I run nequip-evaluate --train-dir /my_path. How to get the predicted values? Thank you very much!
As the title, I get these outputs:

Using device: cpu
Loading model...
loaded model from training session
Loading original dataset...
Loaded validation_dataset specified in config.yaml.
Using origial validation dataset (332 frames) minus validation set frames (332 frames), yielding a test set size of 0 frames.
Starting...
0it [00:00, ?it/s]

--- Final result: ---

[Eipgen]

it seems that yours n_train and n_val are same. if you want to check you training set error, you can follow that.

1. copy a config yaml file and set n_val: 0
2. then add --dataset-config your_copied_config_yaml when use nequip-evaluate
3. also you can use --output and --log for detail results

[phys-chem]

Thank you very much! --output tag gives me what I want. If there is an example for the ASE interface, it will be more convenient.

[Linux-cpp-lisp]

@Eipgen is right that --output is the correct key (you can also use --output-fields if you want to output non-standard fields for analysis, such as the final features, for example). Thanks @Eipgen for responding!

For the ASE interface, see #115, but please note that the ASE interface is not optimal from a performance point of view if you want to run a specific preexisting group of frames, since it cannot batch calculations like nequip-evaluate does.