[Bug]? Not CFG latent_model_input Instead of latents?

[GloryyrolG]

hi @LinB203 @qqingzheng @clownrat6 et al.,

may i check if there is a bug here, where else latents should be else latent_model_input for I2V?

Open-Sora-Plan/opensora/sample/pipeline_inpaint.py

Line 823 in 294993c

latent_model_input = torch.cat([latent_model_input] * 2) if do_classifier_free_guidance else latents

thx & rgds,

[yunyangge]

Thanks for your issue, we will fix this bug.