training_log.json

{
  "base_model_name": "NousResearch_Llama-2-13b-hf",
  "base_model_class": "LlamaForCausalLM",
  "base_loaded_in_4bit": true,
  "base_loaded_in_8bit": false,
  "projections": "q, v",
  "loss": 2.0766,
  "learning_rate": 3.006012024048096e-06,
  "epoch": 1.0,
  "current_steps": 16095,
  "train_runtime": 19428.9967,
  "train_samples_per_second": 3.317,
  "train_steps_per_second": 0.026,
  "total_flos": 6.462072465181901e+17,
  "train_loss": 2.1390227541534848
}