SPB Git

spb/wp31_uqo

Public
999 B · 18 lines yaml
1# Phase 4 step 2: Stage B editor bootstrap on the FROZEN tiny backbone (v1), forced mixed budgets {0,1,2,4,8}.2name: p4_editor_v1_nonzeroU3seed: 14mode: prisme5model: {name: tiny_p4, vocab_size: 128, d_model: 192, n_layers: 4, n_heads: 6, n_kv_heads: 2, ffn_width: 512,6        max_seq_len: 320, w: 64, ffn_w: 192, n_slots: 16, n_editor_heads: 2, r_max: 8}7init: {backbone_checkpoint: experiments/runs/p4_tiny_backbone_seed01_20261029232626/checkpoint, freeze_backbone: true}8data: {kind: synthetic, tokenizer: char, splits_dir: data/splits/synthetic_v1, loss_span: answer,9       eval_splits: [dev_iid, dev_template, dev_depth, dev_compose]}10budgets: [0, 1, 2, 4, 8]11budget_probs: [0.2, 0.2, 0.2, 0.2, 0.2]12budget_per: position13train: {batch_size: 32, seq_len: 320, steps: 2000, lr: 1.0e-3, warmup: 100, weight_decay: 0.1, grad_clip: 1.0,14        log_every: 50, eval_every: 500, eval_batches: 8, ckpt_every: 1500}15em_budgets: [0, 2, 8]16em_max_items: 40017model_overrides: {readout_zero_init: false}18
Files · a2c46f03643fdd75e8e87595931d35e18c590d57
Menu

Theme