| 36 | |
| 37 | |
| 38 | def parse_args() -> argparse.Namespace: |
| 39 | parser = argparse.ArgumentParser( |
| 40 | description="Prepare MOSS-TTS finetuning JSONL by extracting target audio codes." |
| 41 | ) |
| 42 | parser.add_argument("--model-path", type=str, default="OpenMOSS-Team/MOSS-TTS-v1.5") |
| 43 | parser.add_argument("--codec-path", type=str, default="OpenMOSS-Team/MOSS-Audio-Tokenizer") |
| 44 | parser.add_argument( |
| 45 | "--device", |
| 46 | type=str, |
| 47 | default="auto", |
| 48 | help="Audio tokenizer device. Use `auto` to follow the current Accelerate rank device.", |
| 49 | ) |
| 50 | parser.add_argument("--input-jsonl", type=str, required=True) |
| 51 | parser.add_argument("--output-jsonl", type=str, required=True) |
| 52 | parser.add_argument("--batch-size", type=int, default=16) |
| 53 | parser.add_argument("--n-vq", type=int, default=None) |
| 54 | parser.add_argument( |
| 55 | "--num-shards", |
| 56 | type=int, |
| 57 | default=None, |
| 58 | help="Advanced override. Normally inferred from Accelerate world size.", |
| 59 | ) |
| 60 | parser.add_argument( |
| 61 | "--shard-rank", |
| 62 | type=int, |
| 63 | default=None, |
| 64 | help="Advanced override. Normally inferred from Accelerate rank.", |
| 65 | ) |
| 66 | parser.add_argument( |
| 67 | "--skip-reference-audio-codes", |
| 68 | dest="encode_reference_audio", |
| 69 | action="store_false", |
| 70 | help="Skip pre-encoding `ref_audio` / `reference` / `reference_audio`. Default is to encode them.", |
| 71 | ) |
| 72 | parser.add_argument( |
| 73 | "--encode-reference-audio", |
| 74 | dest="encode_reference_audio", |
| 75 | action="store_true", |
| 76 | help=argparse.SUPPRESS, |
| 77 | ) |
| 78 | parser.add_argument( |
| 79 | "--no-encode-reference-audio", |
| 80 | dest="encode_reference_audio", |
| 81 | action="store_false", |
| 82 | help=argparse.SUPPRESS, |
| 83 | ) |
| 84 | parser.add_argument( |
| 85 | "--encode-ref-audio", |
| 86 | dest="encode_reference_audio", |
| 87 | action="store_true", |
| 88 | help=argparse.SUPPRESS, |
| 89 | ) |
| 90 | parser.add_argument( |
| 91 | "--save-shard-suffix", |
| 92 | action="store_true", |
| 93 | help="Advanced override. Save output.rankXXXXX-of-YYYYY.jsonl even in manual sharding mode.", |
| 94 | ) |
| 95 | parser.set_defaults(encode_reference_audio=True) |