Question about implementation of Classifier-Free Guidance

[vim-hjk]

https://github.com/huggingface/diffusers/blob/main/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L984-L986

I understood the formula of the ClASIIFIER-FREE DIFFUSION GUIDANCE paper as below. But I'm confused because the above StableDiffusionPipeline implementation is different. Can someone explain it to me?
$(1+w)ϵ_θ(z_t,c)−wϵ_θ(z_t)=w(ϵ_θ(z_t,c)−ϵ_θ(z_t))+ϵ_θ(z_t,c)$

if self.do_classifier_free_guidance:
    noise_pred_uncond, noise_pred_text = noise_pred.chunk(2)
    # Original code
    # noise_pred = noise_pred_uncond + self.guidance_scale * (noise_pred_text - noise_pred_uncond)
    noise_pred = noise_pred_text + self.guidance_scale * (noise_pred_text - noise_pred_uncond)

[vim-hjk]

$\textnormal{guidance scale} \ s = 1 + w$
$(1+w)ϵ_θ(z_t,c)−wϵ_θ(z_t)=sϵ_θ(z_t,c)-(s - 1)ϵ_θ(z_t)$
$=sϵ_θ(z_t,c)-sϵ_θ(z_t) + ϵ_θ(z_t) = s(ϵ_θ(z_t,c) - ϵ_θ(z_t)) + ϵ_θ(z_t)$
$=ϵ_θ(z_t) + s(ϵ_θ(z_t,c) - ϵ_θ(z_t))$