Ethanfel
becb38c27e
torchaudio was defaulting to the torchcodec backend which requires FFmpeg shared libraries not present in the ComfyUI venv, silently skipping every clip and producing an empty dataset. Also add experiments/vocoder_finetune.json for the BJ vocoder LoRA run (lr=3e-4, rank=128, 10k steps). Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
31 lines
946 B
JSON
31 lines
946 B
JSON
{
|
|
"name": "vocoder_finetune",
|
|
"description": "Single run with fine-tuned BJ BigVGAN vocoder injected. Validates vocoder integration with LoRA training. Best known config: lr=3e-4, rank=128.",
|
|
"data_dir": "/media/unraid/davinci/Selva/BJ/features",
|
|
"output_root": "/media/unraid/davinci/Selva/BJ/experiment/vocoder_finetune",
|
|
"base": {
|
|
"steps": 10000,
|
|
"rank": 128,
|
|
"alpha": 0.0,
|
|
"lr": 3e-4,
|
|
"batch_size": 16,
|
|
"warmup_steps": 200,
|
|
"grad_accum": 1,
|
|
"save_every": 2000,
|
|
"seed": 42,
|
|
"target": "attn.qkv",
|
|
"timestep_mode": "uniform",
|
|
"logit_normal_sigma": 1.0,
|
|
"curriculum_switch": 0.6,
|
|
"lora_dropout": 0.0,
|
|
"lora_plus_ratio": 1.0,
|
|
"lr_schedule": "constant"
|
|
},
|
|
"experiments": [
|
|
{
|
|
"id": "r128_lr_3e4_bj_vocoder",
|
|
"description": "lr=3e-4 rank=128 with fine-tuned BJ BigVGAN vocoder. Direct comparison baseline against previous best g1_r128_lr_3e4."
|
|
}
|
|
]
|
|
}
|