| { | |
| "num_batches": 100000, | |
| "batch_size": 4, | |
| "grad_accum_every": 4, | |
| "learning_rate": 0.001, | |
| "validate_every": 100, | |
| "prime_length": 128, | |
| "generate_every": 500, | |
| "generate_length": 512, | |
| "seq_len": 512, | |
| "use_amp": true, | |
| "use_parametrize": true, | |
| "device": "cuda:0" | |
| } |