{
  "encoder": "encoder",
  "head_layers": 2,
  "max_len": 512,
  "head_max_len": 192,
  "act_costs": {
    "escalate": 0.5
  },
  "amp_dtype": "fp16",
  "model_name": "Phocinae-Largha-150M-v1",
  "temperature": [
    0.8660205006599426,
    0.8081192374229431,
    0.6624661087989807
  ],
  "fine_tuned": true,
  "local_training": {
    "loss": "ce+brier",
    "optimizer": "adafactor",
    "epochs": 1,
    "micro_batch": 8,
    "grad_accum": 4,
    "lr_encoder": 2e-05,
    "lr_head": 0.0001,
    "trainable_params_M": 144.29,
    "updates": 300,
    "hours": 0.044,
    "tag": "n13_r4"
  }
}