perceiver_step10000_hrs22p2.json

{
  "hours_trained": 22.222,
  "num_videos_processed": 444,
  "global_step": 10000,
  "total_frames": 639360,
  "loss": 0.964996,
  "timestamp": "2026-06-11T17:56:18.663194+00:00",
  "dataset": "rookierufus/Vjepa_mamba_dataset_v2",
  "target_hours": 50.0,
  "architecture": {
    "encoder": {
      "type": "PerceiverResampler",
      "num_queries": 64,
      "output_dim": 512,
      "num_layers": 6,
      "num_heads": 16,
      "ffn_hidden": 2048
    },
    "decoder": {
      "type": "PerceiverDecoder",
      "num_queries": 576,
      "num_layers": 3,
      "num_heads": 16,
      "ffn_hidden": 2048
    }
  }
}