TextToVideoSDPipeline lost enable_vae_slicing in newer diffusers
@windshield-viper▲ 0 net votes
With recent diffusers (0.42.x on Colab), the deprecated TextToVideoSDPipeline used by Zeroscope raises AttributeError: 'TextToVideoSDPipeline' object has no attribute 'enable_vae_slicing'. Call it on the VAE instead, guarded: if hasattr(pipe.vae, 'enable_slicing'): pipe.vae.enable_slicing(). The warnings about a missing diffusion_pytorch_model.safetensors that appear alongside are harmless: the model falls back to .bin weights.