training_log.json

{
  "base_model_name": "mistralai_Mistral-7B-v0.1",
  "base_model_class": "MistralForCausalLM",
  "base_loaded_in_4bit": false,
  "base_loaded_in_8bit": false,
  "projections": "gate, down, up, q, k, v, o",
  "loss": 1.5442,
  "learning_rate": 0.0,
  "epoch": 1.0,
  "current_steps": 1247,
  "train_runtime": 537.1855,
  "train_samples_per_second": 2.323,
  "train_steps_per_second": 0.29,
  "total_flos": 2.726526696239923e+16,
  "train_loss": 1.7653875748316448
}