| { |
| "run_name": "chatterbox-am-oss-v1", |
| "loop": "chatterbox_finetuning", |
| "toolkit_dir": "/workspace/chatterbox-finetuning", |
| "git_sha": "df2769b9b639293027beb7138ecb7f20d54b5084", |
| "base_checkpoint": "ResembleAI/chatterbox-multilingual", |
| "warm_start_t3": "/workspace/base-v3/t3_mtl23ls_v3.safetensors", |
| "label_repr": "fidel", |
| "soresa_tts_normalizer": "legacy", |
| "gemination": "off", |
| "label_frontend": "/workspace/gabar-tech/chatterbox-amharic/amharic_text.py:normalize", |
| "label_frontend_sha256": "af9396ee2ba3802bbbc87fab4b95125a02aaaff0a6c7349d812e549686c36580", |
| "base_vocab_size": 2454, |
| "new_vocab_size": 2698, |
| "added_tokens_count": 244, |
| "train_hours": 189.416, |
| "train_clip_count": 41685, |
| "eval_split_index": "/workspace/training/tts-chatterbox/eval/amharic_eval.jsonl", |
| "checkpoint_dir": "/workspace/runs/chatterbox-am-oss-v1/checkpoints", |
| "config": { |
| "paths": { |
| "workdir": "/workspace/runs/chatterbox-am-oss-v1", |
| "manifest": "unused", |
| "eval_split_path": null |
| }, |
| "data": { |
| "quality_floor": 0.6, |
| "languages": [ |
| "am" |
| ], |
| "eval_clip_count": 100, |
| "eval_split_seed": 12648430, |
| "min_duration": 3.0, |
| "max_duration": 10.0, |
| "filter_duration": true, |
| "filter_quality": true, |
| "loudness_normalize": true, |
| "target_dbfs": -23.0, |
| "target_sr": 24000, |
| "dedupe_near_identical": true, |
| "label_quality_max_cer": null, |
| "forced_align_trim": false, |
| "label_repr": "fidel" |
| }, |
| "train": { |
| "base_checkpoint": "ResembleAI/chatterbox-multilingual", |
| "output_name": "chatterbox-am-oss-v1", |
| "use_lora": true, |
| "train_t3": true, |
| "freeze_s3gen": true, |
| "freeze_voice_encoder": true, |
| "lora_r": 64, |
| "lora_alpha": 128, |
| "lora_dropout": 0.05, |
| "learning_rate": 2e-05, |
| "lr_scheduler": "cosine", |
| "warmup_ratio": 0.05, |
| "optimizer": "adamw", |
| "weight_decay": 0.01, |
| "max_grad_norm": 1.0, |
| "precision": "bf16", |
| "gradient_checkpointing": true, |
| "batch_size": 4, |
| "gradient_accumulation_steps": 4, |
| "num_train_epochs": 8, |
| "eval_every_steps": 500, |
| "early_stop_metric": "amharic_cer", |
| "emit_audio_samples": true, |
| "save_steps": 500, |
| "save_total_limit": 1, |
| "logging_steps": 25, |
| "seed": 42 |
| }, |
| "synth": { |
| "sample_rate": 24000, |
| "max_new_tokens": 4096, |
| "temperature": 0.6, |
| "cfg_weight": 0.5, |
| "exaggeration": 0.5, |
| "repetition_penalty": 1.2, |
| "collapse_homophones": true, |
| "gemination": "auto", |
| "label_repr": "fidel", |
| "seed": 1234, |
| "vad_end_trim": true, |
| "min_reference_seconds": 5.0, |
| "max_reference_seconds": 10.0, |
| "warmup_trim_seconds": 0.3, |
| "crossfade_ms": 80, |
| "max_chunk_chars": 260, |
| "pause_final_ms": 220, |
| "pause_emphatic_ms": 180, |
| "pause_intra_ms": 100 |
| }, |
| "eval": { |
| "cer_gate": 0.15, |
| "wer_gate": 0.25, |
| "ecapa_gate": 0.6, |
| "zcr_gate": 0.3, |
| "naturalness_gate": 3.0, |
| "robustness_fail_gate": 0.2, |
| "blind_ab_samples": 30 |
| } |
| }, |
| "base_model": { |
| "repo": "ResembleAI/chatterbox", |
| "revision": "5bb1f6ee58e50c3b8d408bc82a6d3740c2db6e18", |
| "t3_file": "t3_mtl23ls_v3.safetensors", |
| "t3_sha256": "5abca8321ede76f8e61f1cc0d19aea6c946b28871017ce8726f8a69203f05953" |
| } |
| } |
