statistics.json

{
  "total_parameters": 513771207,
  "trainable_parameters": 79388871,
  "frozen_parameters": 434382336,
  "trainable_percentage": 15.452183757740242,
  "vision_total_params": 8787459,
  "vision_trainable_params": 8787459,
  "vision_frozen_params": 0,
  "language_total_params": 494032768,
  "language_trainable_params": 59650432,
  "language_frozen_params": 434382336,
  "adapter_total_params": 5041792,
  "adapter_trainable_params": 5041792,
  "memory_total_params": 5225920,
  "memory_trainable_params": 5225920,
  "vision_4bit_quantized": false,
  "language_4bit_quantized": false,
  "memory_158bit_quantized": true,
  "estimated_model_size_mb": 1026.85,
  "epoch": 0,
  "total_epochs": 10,
  "stage": "stage2",
  "timestamp": "2025-12-19T21:35:56.376020",
  "training_history": [
    {
      "epoch": 0,
      "train_loss": 1.7373690557870052,
      "alignment_loss": 0,
      "learning_rate": 9.786403637412737e-05,
      "gradient_norm": 0
    }
  ]
}