training_metadata.json

{
  "best_val_loss": 1.794412523508072,
  "checkpoint_step": 1000,
  "parameter_count": 2169388,
  "training": {
    "batch_size": 8,
    "block_size": 256,
    "device": "cpu",
    "parameter_count": 2169388,
    "seed": 1337,
    "steps": 1000,
    "threads": 10,
    "torch_version": "2.9.1",
    "variant": "full"
  },
  "training_summary": {
    "batch_size": 8,
    "best_val_loss": 1.794412523508072,
    "block_size": 256,
    "device": "cpu",
    "elapsed_seconds": 1820.8739239999995,
    "parameter_count": 2169388,
    "seed": 1337,
    "steps": 1000,
    "threads": 10,
    "torch_version": "2.9.1",
    "variant": "full"
  },
  "variant": "full"
}