# ARC training config

defaults:
  - arch: hrm_v1
  - _self_

hydra:
  output_subdir: null

# Data path
data_path: data/arc-aug-1000

# Hyperparams - Training
global_batch_size: 768

epochs: 100000
eval_interval: 10000
checkpoint_every_eval: True

lr: 1e-4
lr_min_ratio: 1.0
lr_warmup_steps: 2000

# Standard hyperparameter settings for LM, as used in Llama
beta1: 0.9
beta2: 0.95
weight_decay: 0.1
puzzle_emb_weight_decay: 0.1

# Hyperparams - Puzzle embeddings training
puzzle_emb_lr: 1e-2