https://ouro.foundation/posts/hermes/choosing-the-right-orb-v3-model-for-your-research
looks like we default to orb-v3-direct-inf-omat, and use orb-v3-conservative-inf-omat in the examples. but using the 20 atom cutoff (instead of inf) seems to put serious size limitations on our systems based on RAM. (i seem to max out at 18x18x1 unit cells with 8 atoms per unit cell when using orb-v3-conservative-inf-omat on a v100 with 16GB RAM).
not sure if we can do anything to make this more "plug and play": make memory issues less of a problem, or not make the user need to read up on the potentials?
https://ouro.foundation/posts/hermes/choosing-the-right-orb-v3-model-for-your-research
looks like we default to orb-v3-direct-inf-omat, and use orb-v3-conservative-inf-omat in the examples. but using the 20 atom cutoff (instead of inf) seems to put serious size limitations on our systems based on RAM. (i seem to max out at 18x18x1 unit cells with 8 atoms per unit cell when using orb-v3-conservative-inf-omat on a v100 with 16GB RAM).
not sure if we can do anything to make this more "plug and play": make memory issues less of a problem, or not make the user need to read up on the potentials?