rl_agent_config.json

{
  "encoder": "jhu-clsp/mmBERT-base",
  "head_layers": 2,
  "max_len": 1024,
  "head_max_len": 256,
  "max_prefixes": 6,
  "act_costs": {
    "escalate": 0.5
  },
  "cost_wrong_act": 3.0,
  "amp_dtype": "bf16",
  "model_name": "rl-agent",
  "temperature": [
    0.5,
    1.0,
    1.0
  ],
  "temperature_by_options": {
    "choice:3-5": 0.5
  },
  "training": {
    "description": "Effective V29 sequential fine-tune only; inherited upstream counters omitted.",
    "sequential_chunks": 8,
    "epochs_per_chunk": 1,
    "updates": 23568,
    "elapsed_seconds": 40198
  },
  "fine_tuned": {
    "base_model": "local-checkpoint:devops-v29-p07-r1/model",
    "dataset_sha256": "df4380ef275ce62aff1ba744b1c418278f8bd18e088126de7aded18d9547cb36",
    "best_epoch": 1,
    "created": "2026-10-20T12:59:47",
    "method": "full",
    "objective": "ce",
    "tool": "laya-mlx finetune"
  }
}