{
  "activation_checkpointing": "disabled",
  "commit": "4bc58c2",
  "decision_mode": "trunk",
  "description": "saliencew on lm_head/wte/value_embeds, plain AdamW on scalars, salience gate on Muon, no checkpointing, WARMDOWN_RATIO 0.1, MATRIX_LR 0.05",
  "eval_batch_size": 8,
  "memory_gb": 6.161,
  "num_steps": 38,
  "peak_vram_mb": 6309.3,
  "total_seconds": 453.3,
  "train_batch_size": 8,
  "training_seconds": 309.3,
  "val_bpb": 0.857426
}
