| { | |
| "base_model_name": "TheBloke_rocket-3B-GPTQ", | |
| "base_model_class": "StableLMEpochForCausalLM", | |
| "base_loaded_in_4bit": false, | |
| "base_loaded_in_8bit": false, | |
| "projections": "q, v", | |
| "loss": 1.2437, | |
| "learning_rate": 0.0, | |
| "epoch": 10.0, | |
| "current_steps": 303, | |
| "train_runtime": 106.5088, | |
| "train_samples_per_second": 5.727, | |
| "train_steps_per_second": 0.094, | |
| "total_flos": 160270830796800.0, | |
| "train_loss": 1.4195190906524657 | |
| } |