| { | |
| "dataset_name": "roneneldan/TinyStories", | |
| "max_dataset_samples": 750000, | |
| "tokenizer_vocab_size": 8000, | |
| "context_length": 256, | |
| "d_model": 768, | |
| "n_heads": 12, | |
| "n_layers": 8, | |
| "d_ff": 2304, | |
| "dropout": 0.1, | |
| "batch_size": 32, | |
| "learning_rate": 0.0003, | |
| "weight_decay": 0.1, | |
| "warmup_steps": 2000, | |
| "max_steps": 30000, | |
| "eval_interval": 2000, | |
| "eval_steps": 100, | |
| "save_interval": 5000, | |
| "grad_clip": 1.0, | |
| "device": "cuda", | |
| "fp16": true, | |
| "seed": 42, | |
| "output_dir": "cortex_54m_output" | |
| } |