Downloads · 30 days
0
dnafm/168M_rag_ctx4096_89species_budget80k_best
168M_rag_ctx4096_89species_budget80k_best is a machine learning model from dnafm. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as mit.
⚠️ Requires retrieved MSA input at inference time — do not use as a standalone sequence-only model. This checkpoint was pretrained with retrieval always present (no retrieval dropout), so the model has become dependen…
Downloads · 30 days
0
Access
Public
Updated Aug 25, 2026
Repo size
673 MB
Likes
0
Public
Click a slice to open those files.
.bin673 MB · 100%
From the Hugging Face model README
⚠️ Requires retrieved MSA input at inference time — do not use as a standalone sequence-only model. This checkpoint was pretrained with retrieval always present (no retrieval dropout), so the model has become dependent on the retrieved homologous sequences rather than merely benefiting from them.
remove-gaps or later)import torch
import hydra
from omegaconf import OmegaConf
cfg = OmegaConf.load("network_config.yaml")
network = hydra.utils.instantiate(cfg)
ckpt = torch.load("pytorch_model.bin", map_location="cpu", weights_only=False)
state_dict = {k[len("network."):]: v for k, v in ckpt["state_dict"].items()}
network.load_state_dict(state_dict, strict=True) # verified: 0 missing, 0 unexpected keys
n_retrieved_species): 89species_budget): 80000| Benchmark | Metric | Value |
|---|---|---|
| gene_finding (BEND) | Val MCC | 0.5348 |
| gene_finding (BEND) | Test MCC | 0.5411 |
| var_eff (ClinVar, ctx4096 eval) | AUROC | 0.8702 |
| var_eff (ClinVar, ctx4096 eval) | AUPRC | 0.8179 |