throughput.json

{
  "train_runtime_seconds": 18178.6,
  "train_runtime_hours": 5.05,
  "n_examples": 2271,
  "total_tokens": 10813978,
  "tokens_per_second": 594.9,
  "seconds_per_example": 8.005,
  "mean_tokens_per_example": 4762,
  "peak_vram_gb": 6.78,
  "max_seq_length": 8192,
  "effective_batch": 16,
  "max_steps": 150,
  "gpu": "Tesla T4"
}