Hugging Face's logo

Diffusers documentation

Hybrid Inference API Reference

You are viewing main version, which requires installation from source. If you'd like regular pip install, checkout the latest stable version (v0.32.2).
Hugging Face's logo
Join the Hugging Face community

and get access to the augmented documentation experience

to get started

Hybrid Inference API Reference

Remote Decode

diffusers.utils.remote_decode

< >

( endpoint: str tensor: torch.Tensor processor: typing.Union[ForwardRef('VaeImageProcessor'), ForwardRef('VideoProcessor'), NoneType] = None do_scaling: bool = True scaling_factor: typing.Optional[float] = None shift_factor: typing.Optional[float] = None output_type: typing.Literal['mp4', 'pil', 'pt'] = 'pil' return_type: typing.Literal['mp4', 'pil', 'pt'] = 'pil' image_format: typing.Literal['png', 'jpg'] = 'jpg' partial_postprocess: bool = False input_tensor_type: typing.Literal['binary'] = 'binary' output_tensor_type: typing.Literal['binary'] = 'binary' height: typing.Optional[int] = None width: typing.Optional[int] = None )

Parameters

  • endpoint (str) — Endpoint for Remote Decode.
  • tensor (torch.Tensor) — Tensor to be decoded.
  • processor (VaeImageProcessor or VideoProcessor, optional) — Used with return_type="pt", and return_type="pil" for Video models.
  • do_scaling (bool, default True, optional) — DEPRECATED. pass scaling_factor/shift_factor instead. still set do_scaling=None/do_scaling=False for no scaling until option is removed When True scaling e.g. latents / self.vae.config.scaling_factor is applied remotely. If False, input must be passed with scaling applied.
  • scaling_factor (float, optional) — Scaling is applied when passed e.g. latents / self.vae.config.scaling_factor.

    • SD v1: 0.18215
    • SD XL: 0.13025
    • Flux: 0.3611 If None, input must be passed with scaling applied.
  • shift_factor (float, optional) — Shift is applied when passed e.g. latents + self.vae.config.shift_factor.

    • Flux: 0.1159 If None, input must be passed with scaling applied.
  • output_type ("mp4" or "pil" or "pt", default “pil”) — Endpoint output type. Subject to change. Report feedback on preferred type.

    "mp4": Supported by video models. Endpoint returns bytesof video.“pil”: Supported by image and video models. Image models: Endpoint returns bytesof an image inimage_format. Video models: Endpoint returns torch.Tensorwith partialpostprocessingapplied. Requiresprocessoras a flag (anyNonevalue will work).“pt”: Support by image and video models. Endpoint returns torch.Tensor. With partial_postprocess=Truethe tensor is postprocesseduint8` image tensor.

    Recommendations: "pt" with partial_postprocess=True is the smallest transfer for full quality. "pt" with partial_postprocess=False is the most compatible with third party code. "pil" with image_format="jpg" is the smallest transfer overall.

  • return_type ("mp4" or "pil" or "pt", default “pil”) — Function return type.

    "mp4": Function returns bytesof video.“pil”: Function returns PIL.Image.Image. With output_type=“pil” no further processing is applied. With output_type="pt" a PIL.Image.Imageis created.partial_postprocess=False processoris required.partial_postprocess=True processoris **not** required.“pt”: Function returns torch.Tensor. processoris **not** required.partial_postprocess=Falsetensor isfloat16orbfloat16, without denormalization. partial_postprocess=Truetensor isuint8`, denormalized.

  • image_format ("png" or "jpg", default jpg) — Used with output_type="pil". Endpoint returns jpg or png.
  • partial_postprocess (bool, default False) — Used with output_type="pt". partial_postprocess=False tensor is float16 or bfloat16, without denormalization. partial_postprocess=True tensor is uint8, denormalized.
  • input_tensor_type ("binary", default "binary") — Tensor transfer type.
  • output_tensor_type ("binary", default "binary") — Tensor transfer type.
  • height (int, optional) — Required for "packed" latents.
  • width (int, optional) — Required for "packed" latents.

Hugging Face Hybrid Inference that allow running VAE decode remotely.

< > Update on GitHub

←VAE Decode CogVideoX→