{
  "scope": "Reduced random-weight mechanics; no pretrained visual captions",
  "official_revision": "05cb670f47296522e24e7727e366b4a997d28582",
  "program_sha256": "ee0e457846fe23c1bc519cab544ff767cc89ca257a9d57b1869923e799106e75",
  "command": "python examples/blip2_training.py --smoke --output public/examples/blip2-demo/smoke.json",
  "environment": {
    "python": "3.13.13",
    "torch": "2.12.0+cpu"
  },
  "seed": 17,
  "threads": 2,
  "architecture": {
    "batch": 3,
    "patches": 16,
    "queries": 4,
    "qformer_width": 24,
    "qformer_blocks": 4,
    "lm_width": 32,
    "lm_blocks": 2,
    "vocabulary": 32
  },
  "stage_one": {
    "itc": 1.158509612083435,
    "itm": 0.644074559211731,
    "itg": 3.8968746662139893,
    "loss": 5.699459075927734
  },
  "stage_two_loss": 3.700383186340332,
  "stage_two_labels": [
    [
      -100,
      -100,
      -100,
      -100,
      2,
      4,
      5,
      3,
      -100
    ],
    [
      -100,
      -100,
      -100,
      -100,
      2,
      6,
      7,
      8,
      3
    ],
    [
      -100,
      -100,
      -100,
      -100,
      2,
      9,
      3,
      -100,
      -100
    ]
  ],
  "gradients": {
    "stage_one_qformer": 6.7281494140625,
    "stage_two_qformer": 2.6075737476348877,
    "stage_two_projection": 0.7029546499252319,
    "lm_input_prefix": 0.08519306033849716,
    "frozen_parameters_have_grad": false
  },
  "query_update_norm": 0.016359619796276093,
  "generated_token_ids": [
    [
      2,
      16,
      14,
      25,
      30
    ],
    [
      2,
      16,
      14,
      25,
      30
    ],
    [
      2,
      16,
      14,
      25,
      30
    ]
  ]
}