Skip to content

vllm_omni.diffusion.models.soulx_singer.modules.decoder ¶

CFMDecoder ¶

Bases: Module

model `instance-attribute` ¶

model = FlowMatchingTransformer(cfg=config, **config)

forward ¶

forward(mel, x_mask, decoder_inp, is_prompt)

reverse_diffusion ¶

reverse_diffusion(
    pt_mel,
    pt_decoder_inp,
    gt_decoder_inp,
    n_timesteps=32,
    cfg=1,
)