{ "base_model_name": "TheBloke_rocket-3B-GPTQ", "base_model_class": "StableLMEpochForCausalLM", "base_loaded_in_4bit": false, "base_loaded_in_8bit": false, "projections": "q, v", "train_runtime": 428.7644, "train_samples_per_second": 1.143, "train_steps_per_second": 0.023, "total_flos": 745932954009600.0, "train_loss": 1.526400661468506, "epoch": 5.71, "current_steps": 76, "loss": 1.302, "learning_rate": 0.00023076923076923076 }