{
  "started_at": "2026-01-12T08:52:02+01:00",
  "repos": {
    "merged": "Mathieu-Thomas-JOSSET/joke-20260112-081758",
    "gguf": "Mathieu-Thomas-JOSSET/joke-finetome-model-gguf-phi4-20260112-081758"
  },
  "model_name": "unsloth/Phi-4-unsloth-bnb-4bit",
  "dataset": {
    "name": "Mathieu-Thomas-JOSSET/michael_abab_conversations_infini_instruct.jsonl",
    "split": "train"
  },
  "training": {
    "max_steps": 2000,
    "learning_rate": 9.95267419777795e-06,
    "per_device_train_batch_size": 2,
    "gradient_accumulation_steps": 4,
    "max_seq_length": 2048,
    "seed": 3407,
    "optimizer": "adamw_8bit",
    "lr_scheduler_type": "linear"
  },
  "auto_lr": {
    "enabled": true,
    "use_n": "train",
    "n_ref": 1436,
    "base": 1e-05,
    "mult": 0.5,
    "final": 5e-06
  },
  "metrics": {
    "train_runtime": 2220.807,
    "train_samples_per_second": 5.944,
    "train_steps_per_second": 0.743,
    "total_flos": 4.35104765343744e+16,
    "train_loss": 1.654274252106746,
    "epoch": 3.3342618384401113
  },
  "best": {
    "checkpoint": "/content/outputs/continue_r1_from_350_20260112_073729/checkpoint-100",
    "metric": 2.2380564212799072,
    "metric_name": "eval_loss"
  },
  "plotly": {
    "html": "reports/training_loss_step.html",
    "png": null
  },
  "inference_sample": {
    "source": "dataset",
    "index": 232,
    "messages": [
      {
        "role": "user",
        "content": "Dwight: \"Oh, man.\"\nMichael: \"How did we do it?\"\nDwight: \"I don’t … have no idea.\""
      }
    ]
  }
}