TACE-v1-OAM-M

Version: 0.1.0 Added: 2026-01-06 Published: 2026-01-06 18.8M parameters
Repo Paper Docs DOI Files PR Checkpoint
Predictions
DiscoveryGeometry OptimizationκSRME

Discovery: energy and convex hull diagnostics

Missing preds: 0

Model Authors

  1. Zemin Xu  ShanghaiTech University, Nanjing University  
  2. Wenbo Xie  ShanghaiTech University  
  3. P. Hu  ShanghaiTech University  

Trained By

  1. Zemin Xu  ShanghaiTech University, Nanjing University  

Model Info

  • Version 0.1.0
  • Role Interatomic potential
  • Architecture gnn
  • Targets EFSG
  • Openness OSOD
  • Discovery Train Task S2EFS
  • Discovery Test Task IS2RE-SR

Training Set

MPtrj: 1.58M structures from 146k materials

OMat24: 101M structures from 3.23M materials

Subsampled Alexandria: 10.4M structures from 3.23M materials

description

Based on irreducible Cartesian tensors, described in https://arxiv.org/abs/2509.14961 and https://arxiv.org/abs/2512.16882.

steps

Training performed by: (1) pre-training on OMat24 (lr-4e-3, loss-1:8:8); (2) fine-tuning on MPtrj+sAlex (lr-1e-4, loss-1:1:2).

Hyperparams

  • evaluation: {"max_force":0.05,"max_steps":500,"ase_optimizer":"GOQN","cell_filter":"FrechetCellFilter","kappa":{"protocol":"phonondb-v1","displacement_distance":0.03,"save_forces":true}}
  • architecture: {"graph_construction_radius":6,"n_layers":5}
  • training: {"epochs":4,"initial_learning_rate":0.004,"optimizer":"AdamW","batch_size":512}
  • upstream_config: {"loss_weights":{"energy":1,"force":8,"stress":8},"huber_delta":0.01,"num_channel":48,"Lmax":2,"lmax":3,"l1l2":"l1 <= l2","correlation":2}

Dependencies