ESM-C 6B

5-30 minutesPaperCodeNeed >20GB

ESM-C 6B protein language model

Readiness

Readiness: Need >20GB
Smoke: passed
Artifacts: passed
Production: need_gt20gb_esmc_6b_hf_model_cache_gpu_runtime
Storage: >20GB
Strategy: Use Biohub ESM GitHub package plus Hugging Face cache on larger storage; consider 300M/600M variants for smaller production modes.
Next: Confirm HF access/token policy, provision >20GB model-cache space, install Biohub esm runtime, cache biohub/ESMC-6B or compatible 6B weights, then validate embeddings/logits benchmark through Slurm/API.

2026-07-18T17:33:47+08:00 SOP large-tool batch: reclassified to Need >20GB. Hugging Face lists biohub/esmc-6b-2024-12 as a 6B-parameter F32 model; local full-cache install is expected to exceed 20GB with runtime. Biohub README also shows Hugging Face login in the local example. Backend has no ESM-C 6B cache/runner. Current runner is esm-c-6b-smoke placeholder.