| { | |
| "architecture": "webLTM-recurrent-depth", | |
| "vocab": [ | |
| "0", | |
| "1", | |
| "2", | |
| "3", | |
| "4", | |
| "5", | |
| "6", | |
| "7", | |
| "8", | |
| "9", | |
| "+", | |
| "=", | |
| "ANS" | |
| ], | |
| "seq_len": 18, | |
| "ans_start": 12, | |
| "digits_per_operand": 5, | |
| "d_model": 64, | |
| "n_heads": 4, | |
| "n_blocks": 3, | |
| "block_order": [ | |
| "prelude", | |
| "core", | |
| "coda" | |
| ], | |
| "recurrent_block": "core", | |
| "trained_recurrence_range": [ | |
| 1, | |
| 8 | |
| ], | |
| "task": "non-autoregressive multi-digit addition", | |
| "params": 152909 | |
| } |