| { |
| "config": { |
| "mode": "train", |
| "preset": "qwen36m", |
| "data_dir": "data_final3", |
| "train_file": "", |
| "val_file": "", |
| "vocab_data_dir": "", |
| "max_samples": 0, |
| "max_tokens": 40000000, |
| "val_samples": 1000, |
| "val_batches": 40, |
| "allow_tiny_data": false, |
| "split_tool_segments": true, |
| "init": "qwen", |
| "model_name": "auto", |
| "tokenizer_name": "", |
| "base_layer": 6, |
| "base_layers": 1, |
| "d_model": 896, |
| "n_heads": 14, |
| "d_head": 64, |
| "kv_heads": 2, |
| "intermediate_size": 4864, |
| "attention_bias": true, |
| "rope_theta": 1000000.0, |
| "rms_eps": 1e-06, |
| "depth": 8, |
| "gen_mlp": 128, |
| "factory_prefix": 0, |
| "max_seq_len": 768, |
| "vocab_size": 11264, |
| "vocab_strategy": "coverage", |
| "coverage": 0.999, |
| "merge_weights": "mean", |
| "vocab_pack": "runs/night_glm9q/compact_vocab.pt", |
| "init_ckpt": "runs/night_glm9q/checkpoints/best.pt", |
| "init_skip_embeddings": false, |
| "init_extend_vocab": false, |
| "seed": 42, |
| "batch_size": 4, |
| "grad_accum": 4, |
| "lr": 4e-05, |
| "lr_base": 1e-05, |
| "weight_decay": 0.01, |
| "warmup": 200, |
| "grad_clip": 1.0, |
| "min_lr_ratio": 0.1, |
| "hours": 1.0, |
| "extra_hours": 0.0, |
| "max_steps": 0, |
| "mask": "assistant", |
| "amp": "auto", |
| "device": "auto", |
| "out": "runs/assistant8", |
| "resume": "", |
| "run": "", |
| "student_run": "", |
| "teacher": "Qwen/Qwen2.5-1.5B-Instruct", |
| "teacher_data_teacher": "Qwen/Qwen2.5-3B-Instruct", |
| "teacher_data_backend": "transformers", |
| "teacher_data_url": "http://127.0.0.1:8080/v1/chat/completions", |
| "teacher_data_parallel": 1, |
| "teacher_data_samples": 128, |
| "teacher_data_max_new": 192, |
| "teacher_data_min_accept": 0.8, |
| "teacher_data_use_accepted": false, |
| "teacher_data_seed": 42, |
| "bridge_samples": 128, |
| "bridge_steps": 300, |
| "bridge_max_new": 96, |
| "bridge_batch_size": 2, |
| "bridge_lr": 7e-06, |
| "bridge_lr_base": 1.5e-06, |
| "bridge_save_every": 100, |
| "teacher_device": "auto", |
| "teacher_dtype": "fp16", |
| "df_questions": 64, |
| "df_steps": 600, |
| "df_max_new": 48, |
| "df_rollout_temp": 0.75, |
| "df_top_p": 0.92, |
| "df_kd_temperature": 1.2, |
| "df_hidden_weight": 0.01, |
| "df_anchor_dir": "", |
| "df_anchor_weight": 0.2, |
| "df_lr": 1.5e-06, |
| "df_lr_base": 4e-07, |
| "df_save_every": 100, |
| "eval_seconds": 240, |
| "save_minutes": 15, |
| "sample_seconds": 900, |
| "log_seconds": 10, |
| "no_samples": false, |
| "demo": true, |
| "demo_force": false, |
| "demo_max_new": 64, |
| "export_dtype": "fp16", |
| "compile": false, |
| "grad_checkpoint": false, |
| "opt": "adamw", |
| "muon_lr": 0.02, |
| "tools": true, |
| "max_new": 192, |
| "temperature": 0.0, |
| "rep_penalty": 1.15, |
| "prompt": "", |
| "yes": true, |
| "dry_run": false |
| }, |
| "arch": { |
| "d_model": 896, |
| "attention_bias": true, |
| "rope_theta": 1000000.0, |
| "rms_eps": 1e-06, |
| "n_heads": 14, |
| "d_head": 64, |
| "kv_heads": 2, |
| "intermediate_size": 4864, |
| "base_layers": 1, |
| "depth": 8, |
| "gen_mlp": 128, |
| "factory_prefix": 0, |
| "max_seq_len": 768, |
| "vocab_size": 11264 |
| }, |
| "params": { |
| "g_init": 896, |
| "depth_scale": 8, |
| "base_layer": 14912384, |
| "norm": 896, |
| "embed_tokens": 10092544, |
| "morph_patcher": 158464, |
| "gen_layer": 1792, |
| "factory": 10682368, |
| "TOTAL": 35849352 |
| }, |
| "vocab_meta": { |
| "strategy": "coverage", |
| "vocab_size": 11264, |
| "coverage": 0.9984156092959803, |
| "tokenizer_len": 151665, |
| "d_model": 896, |
| "n_heads": 14, |
| "d_head": 64, |
| "merge_weights": "mean", |
| "pad_old_id": 151643, |
| "created_at": "2026-10-27T17:35:28" |
| }, |
| "created_at": "2026-10-30T16:45:17" |
| } |
