---
library_name: pytorch
tags:
- exafs
- xafs
- materials-science
- spectroscopy
- regression
- moment
base_model: AutonLab/MOMENT-1-small
datasets:
- INnoVerRaL/llm4exafs-training-data
private: true
---

# LLM4EXAFS Baseline Checkpoint

Private release candidate for the baseline model used in the manuscript
*Learning Local Structural Parameters from EXAFS Spectra Using a Pretrained Transformer*.

The model predicts first-shell mean bond length and coordination number from EXAFS spectra.
This checkpoint is the baseline `MOMENT-1-small` + EXAFS + CORAL + energy-embedding run.

## Files

- `model.safetensors`: model-state-only export from
  `checkpoints/baseline_seed_3407/best.pth`.
- `checksums.json`: file size and SHA-256 digest for the uploaded checkpoint.

## Provenance

- Local source checkpoint: `checkpoints/baseline_seed_3407/best.pth`
- Exported format: `safetensors`
- Export scope: `model_state` only, without optimizer or scheduler state
- Best validation objective stored in checkpoint: 0.0582

This is a private draft upload. Public release metadata, citation, and DOI should be
updated after the manuscript data and claims are frozen.