sam-audio-judge

[Linger-fy]

Configured "judge": [
{
"checkpoint_or_model_id": "facebook/sam-audio-judge",
"kind": "judge"
},
1.0
]

Report the following errors：
TypeError: SAMAudioJudgeConfig.init() got an unexpected keyword argument 'dac_vae_encoder'

I don't use sam-audio-judge for sound separation now. The result is very poor. Is this the reason for the poor performance?

[Linger-fy]

My impression is that the model failed to understand the meanings of the descriptions and anchors that I input.

[chevalierNoir]

@Linger-fy

TypeError: SAMAudioJudgeConfig.init() got an unexpected keyword argument 'dac_vae_encoder'

Please check this

I don't use sam-audio-judge for sound separation now. The result is very poor. Is this the reason for the poor performance?

Is it sam-audio or sam-audio-judge? For sam-audio, what's the model (small, base, or large) and have you set predict_span=True and reranking_candidates=8? Can you also describe your scenario, like audio info and text prompt?