| { | |
| "base_model_name": "gpt2", | |
| "base_model_class": "GPT2LMHeadModel", | |
| "base_loaded_in_4bit": false, | |
| "base_loaded_in_8bit": false, | |
| "projections": "c_attn", | |
| "loss": 0.8766, | |
| "learning_rate": 0.00013416407864998736, | |
| "epoch": 15.0, | |
| "current_steps": 677, | |
| "train_runtime": 33.5196, | |
| "train_samples_per_second": 5.37, | |
| "train_steps_per_second": 0.447, | |
| "total_flos": 41438875484160.0, | |
| "train_loss": 0.8892534891764323 | |
| } |