o Gi>@shddlmZddlZddlZddlmZmZddlZddlm Z ddl m Z m Z e rMddl Z ddlmZdd lmZe rBddlZe je je je jd Zdd lmZdBddZ            dCdDd-d.Z    dEdFd1d2Z         dGdHd3d4Z            dCdId6d7Z  dJdKd:d;Z dLdd?Z"  dJdNd@dAZ#dS)O) annotationsN)Literalcast) deprecate)is_safetensors_availableis_torch_available)VaeImageProcessor)VideoProcessor)float16float32bfloat16uint8)ImagedatabytesreturnstrcCsF|drdS|drdS|ds|drdS|dr!d Sd S) NsjpegsPNG  pngsGIF87asGIF89agifsBMbmpunknown) startswith)rrP/home/ubuntu/.local/lib/python3.10/site-packages/diffusers/utils/remote_utils.pydetect_image_type0s   rTpiljpgFbinaryendpointtensor'torch.Tensor' processor-'VaeImageProcessor' | 'VideoProcessor' | None do_scalingboolscaling_factor float | None shift_factor output_typeLiteral['mp4', 'pil', 'pt'] return_type image_formatLiteral['png', 'jpg']partial_postprocessinput_tensor_typeLiteral['binary']output_tensor_typeheight int | NonewidthcCst|jdkr| dur| durtd|dkr&|dkr&| s&t|ttfs&td|r6|dur8tdddd d dSdSdS) Nz1`height` and `width` required for packed latents.ptrz`processor` is required.r&1.0.0zQ`do_scaling` is deprecated, pass `scaling_factor` and `shift_factor` if required.F standard_warn)ndim ValueError isinstancer r r)r!r"r$r&r(r*r+r-r.r0r1r3r4r6rrrcheck_inputs_decode<s"   r?responserequests.Responsec Cs|dks |dkr-|dur-|j}|j}t|d}|d}t|} tjt|| d|}|dkr|rR|dkrJdd|D} t | dkrH| d } | S|dkrP|} | S|dusZ|dkr^|} | St |t rtt t tj|j|dd d } | St tj|j|dd d } | S|dkr|dkr|durtt|jd } t|j} | | _| S|dkr|dur|dkrd d|d d ddddD} | S|dkr|} | S|dkr|dkr|j} | S)Nr8rshapedtyperCcSsg|] }t|qSr)r fromarraynumpy.0imagerrr osz&postprocess_decode..rr)r+RGBcSsg|]}t|qSr)rrErGrrrrJsr r7rmp4)contentheadersjsonloads DTYPE_MAPtorch frombuffer bytearrayreshapelenr>r rlistrpostprocess_video postprocessopenioBytesIOconvertrformatpermutefloatrFroundastype) r@r$r+r-r0 output_tensor parametersrBrC torch_dtypeoutputdetected_formatrrrpostprocess_decode^s`    $ric Cs"i} |||t|jt|jddd} |r|dur|| d<|r)|dur)|| d<|r4|dur4|| d<n|rB|durB|durB|| d<|durR| durR|| d<| | d<d | d <d | d <|d krk|d krk|durkd| d <n|d kr||dkr||dur|d| d <n|dkrd| d <tj|d} | | | dS)N.)r.r+r0rBrCr(r*r&r4r6z tensor/binaryz Content-TypeAcceptrrz image/jpegrz image/pngrMz text/plainr"rparamsrO)rXrBrrCsplit safetensorsrS_tobytes) r"r$r&r(r*r+r.r0r4r6rOre tensor_datarrrprepare_decodes8        rs8Image.Image | list[Image.Image] | bytes | 'torch.Tensor'cCs| dkrtdddddd} | dkrtddd ddd} t|||||||||| | | | | t|||||||| | | d }tj|fi|}|jsMt|t||||| d }|S) aM Hugging Face Hybrid Inference that allow running VAE decode remotely. Args: 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`](https://github.com/huggingface/diffusers/blob/7007febae5cff000d4df9059d9cf35133e8b2ca9/src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion.py#L1083C37-L1083C77). - 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 `bytes` of video. `"pil"`: Supported by image and video models. Image models: Endpoint returns `bytes` of an image in `image_format`. Video models: Endpoint returns `torch.Tensor` with partial `postprocessing` applied. Requires `processor` as a flag (any `None` value will work). `"pt"`: Support by image and video models. Endpoint returns `torch.Tensor`. With `partial_postprocess=True` the tensor is postprocessed `uint8` 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 `bytes` of video. `"pil"`: Function returns `PIL.Image.Image`. With `output_type="pil" no further processing is applied. With `output_type="pt" a `PIL.Image.Image` is created. `partial_postprocess=False` `processor` is required. `partial_postprocess=True` `processor` is **not** required. `"pt"`: Function returns `torch.Tensor`. `processor` is **not** required. `partial_postprocess=False` tensor is `float16` or `bfloat16`, without denormalization. `partial_postprocess=True` tensor is `uint8`, 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. Returns: output (`Image.Image` or `list[Image.Image]` or `bytes` or `torch.Tensor`). base64zinput_tensor_type='base64'r9z9input_tensor_type='base64' is deprecated. Using `binary`.Fr:r zoutput_tensor_type='base64'z:output_tensor_type='base64' is deprecated. Using `binary`.) r"r$r&r(r*r+r.r0r4r6)r@r$r+r-r0) rr?rsrequestspostok RuntimeErrorrPri)r!r"r$r&r(r*r+r-r.r0r1r3r4r6kwargsr@rgrrr remote_decodesn\  r{rI'torch.Tensor' | Image.ImagecCsdS)Nr)r!rIr(r*rrrcheck_inputs_encodeSsr}cCsF|j}|j}t|d}|d}t|}tjt||d|}|S)NrBrCrD) rNrOrPrQrRrSrTrUrV)r@rdrerBrCrfrrrpostprocess_encode\sr~cCsi}i}|dur ||d<|dur||d<t|tjr7tj|d}t|j|d<t|j  dd|d<nt }|j |dd |}|||d S) Nr(r*r"rBrjrkrCPNG)r_rm)r>rSTensorrprq contiguousrXrBrrCror\r]savegetvalue)rIr(r*rOrerbufferrrrprepare_encodehs  rcCsNt||||t|||d}tj|fi|}|js t|t|d}|S)a% Hugging Face Hybrid Inference that allow running VAE encode remotely. Args: endpoint (`str`): Endpoint for Remote Decode. image (`torch.Tensor` or `PIL.Image.Image`): Image to be encoded. scaling_factor (`float`, *optional*): Scaling is applied when passed e.g. [`latents * self.vae.config.scaling_factor`]. - SD v1: 0.18215 - SD XL: 0.13025 - Flux: 0.3611 If `None`, input must be passed with scaling applied. shift_factor (`float`, *optional*): Shift is applied when passed e.g. `latents - self.vae.config.shift_factor`. - Flux: 0.1159 If `None`, input must be passed with scaling applied. Returns: output (`torch.Tensor`). )rIr(r*)r@)r}rrvrwrxryrPr~)r!rIr(r*rzr@rgrrr remote_encode~s$ r)rrrr) NTNNrrrFr r NN)r!rr"r#r$r%r&r'r(r)r*r)r+r,r-r,r.r/r0r'r1r2r3r2r4r5r6r5)NrrF) r@rAr$r%r+r,r-r,r0r') NTNNrrFNN)r"r#r$r%r&r'r(r)r*r)r+r,r.r/r0r'r4r5r6r5)r!rr"r#r$r%r&r'r(r)r*r)r+r,r-r,r.r/r0r'r1r2r3r2r4r5r6r5rrt)NN)r!rrIr|r(r)r*r))r@rA)rIr|r(r)r*r)) r!rrIr|r(r)r*r)rr#)$ __future__rr\rPtypingrrrvdeprecation_utilsr import_utilsrrrSimage_processorr video_processorr safetensors.torchrpr r rrrRPILrrr?rirsr{r}r~rrrrrrs       $ 7 .