| { | |
| "base_model_name": "mpasila_gpt3-finnish-8B-gptq-4bit", | |
| "base_model_class": "BloomForCausalLM", | |
| "base_loaded_in_4bit": false, | |
| "base_loaded_in_8bit": false, | |
| "projections": "query_key_value", | |
| "loss": 2.3115, | |
| "grad_norm": 0.10131204128265381, | |
| "learning_rate": 3.1034482758620685e-05, | |
| "epoch": 2.98, | |
| "current_steps": 1051, | |
| "train_runtime": 4886.022, | |
| "train_samples_per_second": 0.869, | |
| "train_steps_per_second": 0.007, | |
| "total_flos": 112306365136896.0, | |
| "train_loss": 2.4886908531188965 | |
| } |