| { | |
| "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": 0.4541, | |
| "learning_rate": 0.0, | |
| "epoch": 8.0, | |
| "current_steps": 1058, | |
| "train_runtime": 195.5695, | |
| "train_samples_per_second": 7.21, | |
| "train_steps_per_second": 0.051, | |
| "total_flos": 369040514088960.0, | |
| "train_loss": 0.6601485967636108 | |
| } |