| { | |
| "architectures": [ | |
| "FastASR" | |
| ], | |
| "model_type": "blstmp-ctc", | |
| "vocab_size": 2566, | |
| "blank_id": 0, | |
| "unk_id": 1, | |
| "sample_rate": 16000, | |
| "n_mels": 80, | |
| "n_fft": 400, | |
| "win_length": 400, | |
| "hop_length": 160, | |
| "frontend": "Conv2d x4 + MaxPool x2", | |
| "encoder": "4-layer BLSTMP", | |
| "hidden": 320, | |
| "proj": 512, | |
| "rnn_layers": 4, | |
| "dropout": 0.15, | |
| "training": { | |
| "epochs": 25, | |
| "best_epoch": 25, | |
| "optimizer_steps": 178325, | |
| "peak_lr": 0.0006, | |
| "batch": "180s / 90 clips x 4 accum" | |
| } | |
| } |