arkai2025's picture
feat(tools): add voxel-model-eval harness + skill
b7a828f
Raw
History Blame Contribute Delete
412 Bytes
"""Edit this list to add models to evaluate: (label, hf_repo).
The Q4_K_M GGUF filename is resolved at runtime (eval.py picks it from the repo
file list). Hackathon cap: <=32B TOTAL params (MoE counts total, not active).
These downloads are exploration-only and re-downloadable; run
`eval.py --delete-after` to reclaim disk after each model.
"""
MODELS = [
("Gemma4-12B", "unsloth/gemma-4-12b-it-GGUF"),
]