training_log.json

{
  "base_model_name": "LumiOpen_Viking-7B",
  "base_model_class": "LlamaForCausalLM",
  "base_loaded_in_4bit": true,
  "base_loaded_in_8bit": false,
  "projections": "q, v",
  "loss": 1.6425,
  "grad_norm": 0.27980342507362366,
  "learning_rate": 2.9999999999999997e-06,
  "epoch": 1.0,
  "current_steps": 12927,
  "train_runtime": 25093.9843,
  "train_samples_per_second": 2.063,
  "train_steps_per_second": 0.016,
  "total_flos": 1.1247635250680955e+18,
  "train_loss": 1.712549373655036
}