{
  "adapter_sha256": "dcc5f50ea32fe0d59f5dfb1d02fda4996e54a2088618eb622e71f2e0c5dfe693",
  "adapter_step": 162,
  "base_model_id": "Qwen/Qwen2.5-VL-3B-Instruct",
  "levers": {
    "max_pixels": 200704,
    "max_seq_len": 1024,
    "min_pixels": 3136
  },
  "lora": {
    "lora_alpha": 32,
    "lora_dropout": 0.05,
    "lora_r": 16,
    "lora_target_modules": [
      "q_proj",
      "k_proj",
      "v_proj",
      "o_proj",
      "gate_proj",
      "up_proj",
      "down_proj"
    ]
  },
  "optim": {
    "effective_batch_size": 32,
    "epochs": 2,
    "gradient_accumulation_steps": 32,
    "gradient_checkpointing": true,
    "learning_rate_lora": 0.0002,
    "max_grad_norm": 1.0,
    "per_device_batch_size": 1,
    "scheduler": "cosine",
    "seed": 0,
    "warmup_ratio": 0.03
  },
  "quantization": {
    "bnb_4bit_compute_dtype": "bfloat16",
    "bnb_4bit_quant_type": "nf4",
    "bnb_4bit_use_double_quant": true,
    "load_in_4bit": true
  },
  "run_id": "bl_a0ccbcff",
  "stage": "sft_qlora",
  "tied_embedding_status": "tied",
  "tokenizer_version": "vq_v2_srgbres_17to4_cb256_t64__w91cffdd2c82f",
  "train_report": {
    "epochs": 2,
    "mean_loss": 1.9172154127758134,
    "rows": 2864,
    "rows_trained": 5184,
    "seed": 0,
    "skipped": 544,
    "smoke_size": null,
    "steps": 162
  },
  "vocab_size_after_resize": 151924,
  "vq_codebook_sha256": "bcdf369dd7cd9a99d71f240b0dac67d404f52130dc8c35d14d6a04514349d118"
}