Weights for finetuned models?

[jeevster]

Is there any plan to release the weights for the JMP models finetuned on the various data subsets (e.g MD22)? I see that it is possible to finetune from the provided pretrained checkpoints using this repo, but having the checkpoints would be nice to avoid having to do this myself.

[nimashoghi]

I don't think these checkpoints were reserved (as we had 72 different fine-tuning runs, when you take all 5 seeds of MatBench into account), but I'm not 100% sure. @wood-b would have more info on this.

[jeevster]

Ok thanks, please keep me posted. In the meantime, if I do end up finetuning the models myself, I'd like to have the option of finetuning JMP-S in addition to JMP-L. However, I currently only see a config for JMP-L finetuning (src/jmp/configs/finetune/jmp_l.py), and same for the individual datasets (e.g for MD22, there is a jmp_l_md22_config_ in src/jmp/configs/finetune/md22.py but nothing corresponding to JMP-S). Would it be possible to share the JMP-S configs as well?