seed: 1
mode: prisme
model: from_checkpoint
checkpoint: CONTROLLER103M_CKPT
data:
  kind: tokens
  bin: data/processed/fineweb_edu_300m/val.bin
  split: val
batch_size: 8
seq_len: 512
n_batches: 20
budgets:
- 1
- 2
- 4
- 8
oracle_batches: 2
two_step_batches: 1
entropy_thresholds:
- 1.0
- 2.0
- 4.0
norm_thresholds:
- 0.05
- 0.1
- 0.2
lambda_grid:
- 0.0
- 0.03
- 0.1
- 0.3