1experiment:2 name: mini_backbone3 seed: 14 mode: backbone_only5 model:6 name: mini7 vocab_size: 1288 d_model: 969 n_layers: 210 n_heads: 411 n_kv_heads: 212 ffn_width: 25613 max_seq_len: 32014 w: 3215 ffn_w: 9616 n_slots: 1617 n_editor_heads: 218 r_max: 819 readout_zero_init: false20 readout_init_std: 0.221 ctrl_input: h22 ctrl_use_entropy: true23 init:24 freeze_backbone: false25 data:26 kind: synthetic27 tokenizer: char28 splits_dir: data/splits/synthetic_v129 loss_span: all30 eval_splits:31 - dev_iid32 train:33 batch_size: 1634 seq_len: 32035 steps: 15036 lr: 0.00137 warmup: 1038 weight_decay: 0.139 grad_clip: 1.040 log_every: 5041 eval_every: 15042 eval_batches: 243 ckpt_every: 15044 em_max_items: 3245model:46 name: mini47 vocab_size: 12848 d_model: 9649 n_layers: 250 n_heads: 451 n_kv_heads: 252 ffn_width: 25653 max_seq_len: 32054 rope_theta: 10000.055 norm_eps: 1.0e-0556 tie_embeddings: true57 init_std: 0.0258 use_workspace: false59 w: 3260 ffn_w: 9661 n_slots: 1662 n_editor_blocks: 263 n_editor_heads: 264 r_max: 865 eta_max: 0.2566 clip_c: 1.067 clip_eps: 1.0e-0668 readout_zero_init: false69 readout_init_std: 0.270 readout_scale: 1.071 controller_hidden: 072 ctrl_input: h73 ctrl_use_entropy: true74 lambda_cost: 0.075 kappa_risk: 0.076 dense_writes: false77resolved_train:78 batch_size: 1679 seq_len: 32080 steps: 15081 lr: 0.00182 warmup: 1083 weight_decay: 0.184 grad_clip: 1.085 log_every: 5086 eval_every: 15087 eval_batches: 288 ckpt_every: 15089