| { |
| "model_name": "Quantum-Clasifier-Finetuned", |
| "base_model": "Nebulixlabs/Quantum-Classifier", |
| "task": "safe_vs_unsafe", |
| "datasets": [ |
| "Nebulixlabs/safety-dataset", |
| "KimDongH/spam_dataset-train-eval-test2" |
| ], |
| "label_semantics": { |
| "0": "safe", |
| "1": "unsafe" |
| }, |
| "architecture": { |
| "vocab_size": 4096, |
| "context_length": 128, |
| "d_model": 112, |
| "attention_heads": 7, |
| "layers": 4, |
| "ffn": 336, |
| "classifier_hidden": 170, |
| "dropout": 0.1 |
| }, |
| "parameters": 1000000, |
| "batch_size": 128, |
| "target_tokens": 30000000, |
| "actual_tokens": 30010756, |
| "epochs": 9, |
| "optimizer": "AdamW", |
| "learning_rate": 0.0001, |
| "minimum_learning_rate": 1e-05, |
| "weight_decay": 0.01, |
| "warmup_ratio": 0.05, |
| "gradient_clip": 1.0, |
| "mixed_precision": "fp16", |
| "torch_compile": true, |
| "confidence_temperature": 1.4799351692199707, |
| "test_policy": "test split is never used for fine-tuning" |
| } |
