| { | |
| "model": { | |
| "name": "transformer", | |
| "d_model": 256, | |
| "n_layers": 4, | |
| "n_heads": 8, | |
| "d_ff": 1024, | |
| "dropout": 0.0, | |
| "causal": true, | |
| "context": 1024, | |
| "overlap": 256, | |
| "grad_ckpt": false | |
| }, | |
| "train": { | |
| "steps": 10000, | |
| "batch_size": 16, | |
| "seq_len": 1024, | |
| "lr": 0.001, | |
| "min_lr_ratio": 0.1, | |
| "warmup": 1000, | |
| "weight_decay": 0.1, | |
| "grad_clip": 1.0, | |
| "log_every": 50, | |
| "eval_every": 1000, | |
| "eval_batch_tokens": 16384, | |
| "seed": 0, | |
| "align_crops": true | |
| }, | |
| "params": 3184156, | |
| "data": { | |
| "pairs": 569670, | |
| "dropped": 0, | |
| "null_symbols": "!'+,-./:;<=>?[]_{}" | |
| }, | |
| "best": { | |
| "char_accuracy": 0.1778054242826548, | |
| "word_accuracy": 0.007266449651063734, | |
| "precision": 0.78400927230149, | |
| "recall": 0.046872386788544516, | |
| "f1": 0.030376188426197034, | |
| "levenshtein": 742.7077189939289, | |
| "normalized_levenshtein": 0.822043863867842, | |
| "samples": 1153, | |
| "loss": 2.5003384844661993, | |
| "step": 10000 | |
| } | |
| } |