{ "base_model": "Qwen/Qwen3.5-9B", "vev_version": "0.1.0", "train_args": { "build": "v1-research", "base": "Qwen/Qwen3.5-9B", "out": "train", "limit": 100000, "sources": null, "source_weights": null, "max_steps": 2500, "lr": 5e-05, "head_lr": 0.0, "lora_r": 16, "lora_alpha": 0, "lora_dropout": 0.05, "lora_targets": "all", "dp": 256, "prior": true, "anchor_kl": 0.3, "anchor_build": "anchor-v3", "anchor_weight": 2.0, "anchor_sym": false, "consistency": 0.0, "consistency_frac": 0.5, "micro_tokens": 16384, "accum": 4, "max_q": 4, "max_row_tokens": 4096, "max_pixels": 1048576, "brier": 0.1, "score_w": 0.05, "weight_decay": 0.0, "grad_clip": 1.0, "warmup_pct": 0.1, "no_augment": false, "no_gradient_checkpointing": false, "dtype": "bf16", "attn": null, "workers": 8, "eval_every": 250, "eval_rows": 1000, "eval_records": 2000, "save_every": 250, "log_every": 10, "seed": 0, "resume": null, "device": "cuda" }, "base": "." }