training_log.json

{
  "base_model_name": "meta-llama_Meta-Llama-3-8B",
  "base_model_class": "LlamaForCausalLM",
  "base_loaded_in_4bit": false,
  "base_loaded_in_8bit": false,
  "projections": "q, v",
  "loss": 0.9512,
  "grad_norm": 0.6213752031326294,
  "learning_rate": 0.0005979381443298969,
  "epoch": 4.028685258964144,
  "current_steps": 10055,
  "current_steps_adjusted": 10055,
  "epoch_adjusted": 4.028685258964144,
  "train_runtime": 3864.6216,
  "train_samples_per_second": 25.977,
  "train_steps_per_second": 0.101,
  "total_flos": 4.69608070712918e+17,
  "train_loss": 1.8759732170950008
}