| { | |
| "architecture": "DiffusionLM", | |
| "T": 200, | |
| "d_model": 256, | |
| "n_layers": 4, | |
| "d_ff": 1024, | |
| "n_heads": 4, | |
| "a_len": 4, | |
| "q_len": 28, | |
| "noise_schedule": "sqrt_power_law", | |
| "clamping": "euclidean", | |
| "loss": "mse_ce", | |
| "decoder_causal_mask": false, | |
| "vocab_size": 516, | |
| "params": 8024576, | |
| "training_checkpoint": "Phase 3 Reconstructed", | |
| "training_steps": 500, | |
| "training_lr": 0.0003, | |
| "training_seed": 42, | |
| "dataset": "prajwaldongre/riddles-a-synthetic-riddle-dataset-for-nlp", | |
| "train_exact_match": 87.5, | |
| "heldout_exact_match": 47.0, | |
| "heldout_k10_exact_match": 47.0 | |
| } | |