audio8-tts-lora-finetuning/instavar-voice-backend.json at main · instavar/audio8-tts-lora-finetuning · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
{
  "schema_version": "1.2.0",
  "backend_id": "audio8-tts-lora-pytorch",
  "capability_binding": {
    "manifest": "instavar-voice-capabilities.json",
    "adaptation": "lora",
    "runtime_ids": ["pytorch_cuda", "pytorch_mps"]
  },
  "required_environment": [
    { "name": "RAW_TRAIN_JSONL", "purpose": "Raw training split used for grouped corpus audit." },
    { "name": "RAW_VALIDATION_JSONL", "purpose": "Raw validation split used for grouped corpus audit." },
    { "name": "RAW_TEST_JSONL", "purpose": "Raw test split used for grouped corpus audit." },
    { "name": "TRAIN_JSONL", "purpose": "Prepared Audio8 codec-index training manifest." },
    { "name": "EVAL_JSONL", "purpose": "Prepared Audio8 codec-index validation manifest." },
    { "name": "PREPARED_TRAIN_ROOT", "purpose": "Root containing the training manifest and every referenced codec artifact." },
    { "name": "PREPARED_VALIDATION_ROOT", "purpose": "Root containing the validation manifest and every referenced codec artifact." },
    { "name": "DATASET_LINEAGE", "purpose": "Content-addressed receipt binding audited splits to both prepared codec trees." },
    { "name": "BASE_MODEL_DIR", "purpose": "Pinned local Audio8 base-model directory." },
    { "name": "REFERENCE_AUDIO", "purpose": "Authorized reference-speaker audio for reload and held-out synthesis." },
    { "name": "REFERENCE_TEXT", "purpose": "Transcript matching the authorized reference audio." },
    { "name": "SELECTED_ADAPTER_NAME", "purpose": "Single safe Trainer checkpoint directory selected for reload." },
    { "name": "GENERATION_PLAN", "purpose": "Frozen Instavar Voice generation plan containing this candidate." },
    { "name": "CANDIDATE_ID", "purpose": "Candidate identifier present in the frozen generation plan." },
    { "name": "DEVICE", "purpose": "Explicit cuda or mps runtime used by this lifecycle run." },
    { "name": "DTYPE", "purpose": "Explicit inference precision compatible with the selected runtime." },
    { "name": "INSTAVAR_VOICE_EVALUATOR_DIR", "purpose": "Pinned local Instavar Voice evaluator checkout used for required objective extractors." },
    { "name": "FASTER_WHISPER_PYTHON", "purpose": "Executable from the isolated faster-whisper environment." },
    { "name": "FASTER_WHISPER_MODEL_DIR", "purpose": "Symlink-free local faster-whisper model tree." },
    { "name": "FASTER_WHISPER_MODEL_NAME", "purpose": "Declared faster-whisper model identifier." },
    { "name": "FASTER_WHISPER_MODEL_REVISION", "purpose": "Exact faster-whisper model revision." },
    { "name": "FASTER_WHISPER_DEVICE", "purpose": "Explicit device for ASR evaluation." },
    { "name": "FASTER_WHISPER_COMPUTE_TYPE", "purpose": "Explicit CTranslate2 compute type for ASR evaluation." },
    { "name": "SPEECHBRAIN_PYTHON", "purpose": "Executable from the isolated SpeechBrain environment." },
    { "name": "SPEECHBRAIN_MODEL_DIR", "purpose": "Symlink-free local SpeechBrain ECAPA model tree." },
    { "name": "SPEECHBRAIN_MODEL_REVISION", "purpose": "Exact SpeechBrain ECAPA model revision." },
    { "name": "SPEECHBRAIN_DEVICE", "purpose": "Explicit device for speaker evaluation." },
    { "name": "SPEECHBRAIN_TRUST_MODEL_CHECKPOINTS", "purpose": "Explicit true acknowledgement for trusted local SpeechBrain checkpoint loading." },
    { "name": "SPEAKER_REFERENCE_ID", "purpose": "Frozen target-speaker reference identifier." },
    { "name": "SPEAKER_REFERENCE_CATALOG_ID", "purpose": "Frozen target-speaker catalog identifier." },
    { "name": "SPEAKER_REFERENCE_CATALOG", "purpose": "Preregistered content-addressed speaker-reference catalog." },
    { "name": "SPEAKER_REFERENCE_TRANSCRIPT", "purpose": "Transcript file matching REFERENCE_AUDIO." },
    { "name": "SPEAKER_REFERENCE_PLAN", "purpose": "Preregistered speaker-reference assignment plan bound to GENERATION_PLAN." },
    { "name": "PERSISTED_PACKAGE_ROOT", "purpose": "Existing durable directory outside the checkout and lifecycle work directory for the content-addressed final package." }
  ],
  "commands": {
    "preflight": ["{python}", "scripts/instavar_voice_lifecycle.py", "preflight"],
    "train": ["{python}", "scripts/instavar_voice_lifecycle.py", "train"],
    "infer": ["{python}", "scripts/instavar_voice_lifecycle.py", "infer"],
    "evaluate": ["{python}", "scripts/instavar_voice_lifecycle.py", "evaluate"],
    "package": ["{python}", "scripts/instavar_voice_lifecycle.py", "package"]
  },
  "expected_artifacts": {
    "preflight": ["preflight/preflight.json"],
    "train": ["train/selected-adapter.tar"],
    "infer": ["infer/candidate.wav"],
    "evaluate": ["evaluate/evaluation-bundle.tar"],
    "package": ["package/adapter-package.tar", "package/persisted-package.json"]
  },
  "timeout_seconds": {
    "preflight": 1800,
    "train": 86400,
    "infer": 3600,
    "evaluate": 21600,
    "package": 1800
  }
}