| { | |
| "checkpoint": "ckpt_110.pt", | |
| "model": "GPT2-M", | |
| "training_step": 78000, | |
| "instability_type": "Recoverable loss spikes", | |
| "learning_rate": "3e-3", | |
| "decay": "0.1", | |
| "warm": "2000", | |
| "data_type": "BF16", | |
| "reproducibility": false | |
| } |
| { | |
| "checkpoint": "ckpt_110.pt", | |
| "model": "GPT2-M", | |
| "training_step": 78000, | |
| "instability_type": "Recoverable loss spikes", | |
| "learning_rate": "3e-3", | |
| "decay": "0.1", | |
| "warm": "2000", | |
| "data_type": "BF16", | |
| "reproducibility": false | |
| } |