WARNING: Using multiheads finetuning with a foundation model that is not a Materials Project model

[liubaoshuai1402]

Dear developers,

I try to use multiheads finetuning with a foundation model, mace_agnesi_medium.model or mace_agnesi_small.model.
But MACE 0.39 seems don't recognize them as a MP model and returns a WARNING in the log: Using multiheads finetuning with a foundation model that is not a Materials Project model, need to provied a path to a pretraining file with --pt_train_file.
Here is my input YAML, I wonder whether I have neglected some important parameters to open multiheads finetuning?
input.txt
Any advice will be appreciated, thanks!

[Ssbban]

Hi,
I am facing the same issue. Could you please give me some advice how you resolved it? Thanks for your help.

[Ssbban]

Hi Ilyes, thanks for your response. I found that the example on the mace user guide web did not use pt_train_file while automatically identified it as a Material Project model. What causes this difference?

mace_run_train \
    --name="MACE" \
    --foundation_model="mace_agnesi_small.model" \
    --multiheads_finetuning=True \
    --train_file="train.xyz" \
    --valid_fraction=0.05 \
    --test_file="test.xyz" \
    --energy_weight=1.0 \
    --forces_weight=1.0 \
    --E0s="{"1": 130.0 }" \
    --lr=0.01 \
    --scaling="rms_forces_scaling" \
    --batch_size=2 \
    --max_num_epochs=6 \
    --ema \
    --ema_decay=0.99 \
    --amsgrad \
    --default_dtype="float64" \
    --device=cuda \
    --seed=3

[ilyes319]

new version i did not update the doc yet.

[Ssbban]

Hi Ilyes, thank you for the clarification, but I am currently using version 0.3.8 of MACE. Could this version require any additional adjustments to the input settings?

[liubaoshuai1402]

Hi Ilyes, thank you for the clarification, but I am currently using version 0.3.8 of MACE. Could this version require any additional adjustments to the input settings?

mace_run_train -h will output all the related input key words and its meaning

[Ssbban]

Thanks