configs/ablation.yaml

# ─────────────────────────────────────────────────────────────
# configs/ablation.yaml
# Phase 1: lambda_kl ablation on a 10k subset.
# Run FIRST before any full training.
# ─────────────────────────────────────────────────────────────

defaults:
  - base

experiment_name : "ablation"

data:
  n_samples : 10000    # ablation uses 10k for speed

training:
  epochs : 15          # sufficient to converge on 10k

scheduler:
  T_max : 15

early_stopping:
  patience : 5

wandb:
  log_attention_every_n_epochs : 99   # disable attention in ablation