training_log.json

{
  "base_model_name": "Llama-2-7b-chat-hf",
  "base_model_class": "LlamaForCausalLM",
  "base_loaded_in_4bit": true,
  "base_loaded_in_8bit": false,
  "projections": "q, v",
  "loss": 0.8314,
  "learning_rate": 7.264545643486997e-07,
  "epoch": 2.76,
  "current_steps": 3387,
  "train_runtime": 13542.7743,
  "train_samples_per_second": 0.277,
  "train_steps_per_second": 0.002,
  "total_flos": 1.4593684047239578e+17,
  "train_loss": 0.9451417039941858
}