You can deploy this model and start earning money today!

zxcghoul3228/
DOSSIER

Name: DOSSIER
Availability: InStock

text-embeddings-inference

endpoints_compatible

Views

No views yet

DOSSIER

DOSSIER (Density of States from Stoichiometry with Encoder Representations) maps a stoichiometric chemical formula to the electronic density of states (DOS) without requiring a crystal structure.

Encoder: ModernBERT (pretrained by distillation from a universal ML interatomic potential)
Decoder: transposed-convolution head onto a 401-point energy grid (−10 to +10 eV vs. Fermi level)
Units: states eV⁻¹ atom⁻¹

Demo Space: zxcghoul3228/DOSSIER_demo

Files

dos_model.pth — full inference state dict (encoder + decoder)
config.json / tokenizer files — ModernBERT tokenizer and config

Inference

Load config.json with AutoModel.from_config, wrap with the deconv head, then load_state_dict from dos_model.pth. See the Space inference.py.