training_logs/github_actions/training_log_here.json

{
  "model": "6.8M ternary",
  "vocab_size": 1024,
  "dim": 256,
  "layers": 4,
  "steps": 50,
  "batch_size": 4,
  "seq_len": 256,
  "initial_loss": 6.948805332183838,
  "final_loss": 6.937704563140869,
  "loss_drop": 0.01110076904296875,
  "time_seconds": 23.393130779266357,
  "fp16_mb": 13.008331298828125,
  "ternary_mb": 1.2845727157592775,
  "compression": 10.1,
  "swap": "14GB (10+5) active",
  "ram": "1.9GB",
  "hardware": "Consumer hardware only, no data center",
  "method_correct": true,
  "credits": "deepRcurs Labs @deeprcurs / Mzed Imamkh @mzedimamkh"
}