推理时遇到的问题

[ikunikun-ass]

ValueError: The following model_kwargs are not used by the model: ['audios', 'audio_indices'] (note: typos in the generate arguments will also show up in this list)在inference_sts中的
outputs = self.model.generate(
input_ids,
audios=audios,
audio_indices=audio_indices,
)
会报错，在注释掉audios和audio_indices后又报错
RuntimeError: The expanded size of the tensor (113) must match the existing size (14) at non-singleton dimension 3. Target sizes: [1, 28, 14, 113]. Tensor sizes: [1, 1, 14, 14]
使用的是vita_audio_boost，求助

[jordanqin119]

same issues，maybe qwen2.5 is a prune text model, can't accept audio input

[jtyissix]

use flashattention