{
  "base_model": "Qwen/Qwen3.8-27B",
  "base_revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
  "best_checkpoint": "/tmp/grug-v2-lora/checkpoint-600",
  "best_validation_loss": 0.3083864152431488,
  "dataset_receipt": {
    "backbone_model": "Qwen/Qwen3.8-27B",
    "backbone_revision": "1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0",
    "groups_disjoint": true,
    "loss_mask": "last assistant body, reasoning, answer/tool call, EOS; all prompts and history masked",
    "max_length": 6144,
    "packing": false,
    "source_revision": "9bdedcdf86c3b4618f6526879282a5532ee8c690",
    "splits": {
      "train": 3466,
      "validation": 176
    },
    "statistics": {
      "overlength_agentic_continue": 86,
      "overlength_agentic_recovery": 7,
      "train/agentic_continue/low": 185,
      "train/agentic_continue/medium": 213,
      "train/agentic_continue/xhigh": 183,
      "train/agentic_recovery/low": 65,
      "train/agentic_recovery/medium": 65,
      "train/agentic_recovery/xhigh": 65,
      "train/chat/low": 29,
      "train/chat/medium": 20,
      "train/chat/xhigh": 28,
      "train/coding/low": 269,
      "train/coding/medium": 302,
      "train/coding/xhigh": 219,
      "train/general_qa/low": 24,
      "train/general_qa/medium": 11,
      "train/general_qa/xhigh": 23,
      "train/identity_format/low": 9,
      "train/identity_format/medium": 10,
      "train/identity_format/xhigh": 20,
      "train/math/low": 307,
      "train/math/medium": 311,
      "train/math/xhigh": 270,
      "train/title/low": 92,
      "train/title/medium": 92,
      "train/title/xhigh": 92,
      "train/verified_repository_repair/low": 188,
      "train/verified_repository_repair/medium": 155,
      "train/verified_repository_repair/xhigh": 219,
      "train_supervised_tokens": 968945,
      "train_tokens": 4476496,
      "validation/agentic_continue/low": 13,
      "validation/agentic_continue/medium": 9,
      "validation/agentic_continue/xhigh": 11,
      "validation/agentic_recovery/low": 5,
      "validation/agentic_recovery/medium": 3,
      "validation/agentic_recovery/xhigh": 10,
      "validation/chat/low": 2,
      "validation/chat/xhigh": 1,
      "validation/coding/low": 17,
      "validation/coding/medium": 20,
      "validation/coding/xhigh": 16,
      "validation/general_qa/low": 1,
      "validation/general_qa/medium": 1,
      "validation/identity_format/medium": 1,
      "validation/math/low": 12,
      "validation/math/medium": 15,
      "validation/math/xhigh": 12,
      "validation/title/low": 4,
      "validation/title/medium": 4,
      "validation/title/xhigh": 4,
      "validation/verified_repository_repair/low": 5,
      "validation/verified_repository_repair/xhigh": 10,
      "validation_supervised_tokens": 61905,
      "validation_tokens": 269482
    },
    "template_sha256": "cd6193bb4cdc3ee69abd2ae9e15dc05ec3b85a48d161a623bdaae54009c0f995",
    "tokenizer_source_revision": "3ab073b4bb06dc8a83e819a33485469f32b945ba",
    "vocabulary_matches_base": true,
    "vocabulary_sha256": "365c2d8a0c0e72d4bcc5dbc6c2e9330a87dae7ffc77e9fcd9d65739f4d16bfff"
  },
  "epochs_requested": 1.5,
  "global_steps": 651,
  "hub_kernels_enabled": true,
  "initial_validation": {
    "epoch": 0,
    "eval_loss": 0.7974646091461182,
    "eval_model_preparation_time": 0.0405,
    "eval_runtime": 76.2112,
    "eval_samples_per_second": 1.26,
    "eval_steps_per_second": 1.26
  },
  "learning_rate": 2e-05,
  "liger_fused_linear_cross_entropy": true,
  "method": "LoRA r=32 alpha=64",
  "runtime": {
    "cuda": "13.0",
    "gpu": "NVIDIA H200",
    "peft": "0.20.0",
    "torch": "2.13.0+cu130",
    "transformers": "5.16.1"
  },
  "target_modules_regex": "model\\.language_model\\.layers\\.\\d+\\.(?:self_attn\\.(?:q_proj|k_proj|v_proj|o_proj)|linear_attn\\.(?:in_proj_qkv|in_proj_a|in_proj_b|in_proj_z|out_proj)|mlp\\.(?:gate_proj|up_proj|down_proj))",
  "train_metrics": {
    "epoch": 1.5005767012687428,
    "step": 651,
    "total_flos": 1.1746914808283873e+18,
    "train_loss": 0.36738806399881563,
    "train_runtime": 4282.0163,
    "train_samples_per_second": 1.214,
    "train_steps_per_second": 0.152
  },
  "trainable_parameters": 233455616,
  "training_max_sequence_length": 6144,
  "training_sequences": 3466
}