feat(api): add img2img mode to panorama pipeline

api/onnx_web/diffusers/pipelines/panorama.py

@@ -15,13 +15,14 @@
 import inspect
 from typing import Callable, List, Optional, Union
 
+import PIL
 import numpy as np
 import torch
 from transformers import CLIPImageProcessor, CLIPTokenizer
 
 from diffusers.configuration_utils import FrozenDict
 from diffusers.schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
-from diffusers.utils import deprecate, logging
+from diffusers.utils import deprecate, logging, PIL_INTERPOLATION
 from diffusers.pipelines.onnx_utils import ORT_TO_NP_TYPE, OnnxRuntimeModel
 from diffusers.pipeline_utils import DiffusionPipeline
 from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
@@ -30,6 +31,27 @@ from diffusers.pipelines.stable_diffusion import StableDiffusionPipelineOutput
 logger = logging.get_logger(__name__)
 
 
+def preprocess(image):
+    if isinstance(image, torch.Tensor):
+        return image
+    elif isinstance(image, PIL.Image.Image):
+        image = [image]
+
+    if isinstance(image[0], PIL.Image.Image):
+        w, h = image[0].size
+        w, h = (x - x % 64 for x in (w, h))  # resize to integer multiple of 64
+
+        image = [np.array(i.resize((w, h), resample=PIL_INTERPOLATION["lanczos"]))[None, :] for i in image]
+        image = np.concatenate(image, axis=0)
+        image = np.array(image).astype(np.float32) / 255.0
+        image = image.transpose(0, 3, 1, 2)
+        image = 2.0 * image - 1.0
+        image = torch.from_numpy(image)
+    elif isinstance(image[0], torch.Tensor):
+        image = torch.cat(image, dim=0)
+    return image
+
+
 class OnnxStableDiffusionPanoramaPipeline(DiffusionPipeline):
     vae_encoder: OnnxRuntimeModel
     vae_decoder: OnnxRuntimeModel
@@ -491,3 +513,236 @@ class OnnxStableDiffusionPanoramaPipeline(DiffusionPipeline):
             return (image, has_nsfw_concept)
 
         return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)
+
+    def img2img(
+        self,
+        prompt: Union[str, List[str]] = None,
+        image: Union[np.ndarray, PIL.Image.Image] = None,
+        strength: float = 0.8,
+        num_inference_steps: Optional[int] = 50,
+        guidance_scale: Optional[float] = 7.5,
+        negative_prompt: Optional[Union[str, List[str]]] = None,
+        num_images_per_prompt: Optional[int] = 1,
+        eta: Optional[float] = 0.0,
+        generator: Optional[np.random.RandomState] = None,
+        prompt_embeds: Optional[np.ndarray] = None,
+        negative_prompt_embeds: Optional[np.ndarray] = None,
+        output_type: Optional[str] = "pil",
+        return_dict: bool = True,
+        callback: Optional[Callable[[int, int, np.ndarray], None]] = None,
+        callback_steps: int = 1,
+    ):
+        r"""
+        Function invoked when calling the pipeline for generation.
+
+        Args:
+            prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts to guide the image generation. If not defined, one has to pass `prompt_embeds`.
+                instead.
+            image (`PIL.Image.Image` or List[`PIL.Image.Image`] or `torch.FloatTensor`):
+                `Image`, or tensor representing an image batch which will be upscaled. *
+            num_inference_steps (`int`, *optional*, defaults to 50):
+                The number of denoising steps. More denoising steps usually lead to a higher quality image at the
+                expense of slower inference.
+            guidance_scale (`float`, *optional*, defaults to 7.5):
+                Guidance scale as defined in [Classifier-Free Diffusion Guidance](https://arxiv.org/abs/2207.12598).
+                `guidance_scale` is defined as `w` of equation 2. of [Imagen
+                Paper](https://arxiv.org/pdf/2205.11487.pdf). Guidance scale is enabled by setting `guidance_scale >
+                1`. Higher guidance scale encourages to generate images that are closely linked to the text `prompt`,
+                usually at the expense of lower image quality.
+            negative_prompt (`str` or `List[str]`, *optional*):
+                The prompt or prompts not to guide the image generation. If not defined, one has to pass
+                `negative_prompt_embeds`. instead. Ignored when not using guidance (i.e., ignored if `guidance_scale`
+                is less than `1`).
+            num_images_per_prompt (`int`, *optional*, defaults to 1):
+                The number of images to generate per prompt.
+            eta (`float`, *optional*, defaults to 0.0):
+                Corresponds to parameter eta (η) in the DDIM paper: https://arxiv.org/abs/2010.02502. Only applies to
+                [`schedulers.DDIMScheduler`], will be ignored for others.
+            generator (`np.random.RandomState`, *optional*):
+                One or a list of [numpy generator(s)](TODO) to make generation deterministic.
+            latents (`np.ndarray`, *optional*):
+                Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
+                generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
+                tensor will ge generated by sampling using the supplied random `generator`.
+            prompt_embeds (`np.ndarray`, *optional*):
+                Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
+                provided, text embeddings will be generated from `prompt` input argument.
+            negative_prompt_embeds (`np.ndarray`, *optional*):
+                Pre-generated negative text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt
+                weighting. If not provided, negative_prompt_embeds will be generated from `negative_prompt` input
+                argument.
+            output_type (`str`, *optional*, defaults to `"pil"`):
+                The output format of the generate image. Choose between
+                [PIL](https://pillow.readthedocs.io/en/stable/): `PIL.Image.Image` or `np.array`.
+            return_dict (`bool`, *optional*, defaults to `True`):
+                Whether or not to return a [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] instead of a
+                plain tuple.
+            callback (`Callable`, *optional*):
+                A function that will be called every `callback_steps` steps during inference. The function will be
+                called with the following arguments: `callback(step: int, timestep: int, latents: torch.FloatTensor)`.
+            callback_steps (`int`, *optional*, defaults to 1):
+                The frequency at which the `callback` function will be called. If not specified, the callback will be
+                called at every step.
+
+        Returns:
+            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] or `tuple`:
+            [`~pipelines.stable_diffusion.StableDiffusionPipelineOutput`] if `return_dict` is True, otherwise a `tuple.
+            When returning a tuple, the first element is a list with the generated images, and the second element is a
+            list of `bool`s denoting whether the corresponding generated image likely represents "not-safe-for-work"
+            (nsfw) content, according to the `safety_checker`.
+        """
+
+        height = image.height
+        width = image.width
+
+        # check inputs. Raise error if not correct
+        self.check_inputs(
+            prompt, height, width, callback_steps, negative_prompt, prompt_embeds, negative_prompt_embeds
+        )
+
+        # define call parameters
+        if prompt is not None and isinstance(prompt, str):
+            batch_size = 1
+        elif prompt is not None and isinstance(prompt, list):
+            batch_size = len(prompt)
+        else:
+            batch_size = prompt_embeds.shape[0]
+
+        if generator is None:
+            generator = np.random
+
+        # here `guidance_scale` is defined analog to the guidance weight `w` of equation (2)
+        # of the Imagen paper: https://arxiv.org/pdf/2205.11487.pdf . `guidance_scale = 1`
+        # corresponds to doing no classifier free guidance.
+        do_classifier_free_guidance = guidance_scale > 1.0
+
+        prompt_embeds = self._encode_prompt(
+            prompt,
+            num_images_per_prompt,
+            do_classifier_free_guidance,
+            negative_prompt,
+            prompt_embeds=prompt_embeds,
+            negative_prompt_embeds=negative_prompt_embeds,
+        )
+
+        # get the initial random noise unless the user supplied it
+        latents_dtype = prompt_embeds.dtype
+
+        # set timesteps
+        self.scheduler.set_timesteps(num_inference_steps)
+
+        # prep image
+        image = preprocess(image).cpu().numpy()
+        image = image.astype(latents_dtype)
+        # encode the init image into latents and scale the latents
+        latents = self.vae_encoder(sample=image)[0]
+        latents = 0.18215 * latents
+
+        latents = latents * np.float64(self.scheduler.init_noise_sigma)
+
+        # get the original timestep using init_timestep
+        offset = self.scheduler.config.get("steps_offset", 0)
+        init_timestep = int(num_inference_steps * strength) + offset
+        init_timestep = min(init_timestep, num_inference_steps)
+
+        timesteps = self.scheduler.timesteps.numpy()[-init_timestep]
+        timesteps = np.array([timesteps] * batch_size * num_images_per_prompt)
+
+        noise = generator.randn(*init_latents.shape).astype(latents_dtype)
+        init_latents = self.scheduler.add_noise(
+            torch.from_numpy(init_latents), torch.from_numpy(noise), torch.from_numpy(timesteps)
+        )
+        init_latents = init_latents.numpy()
+
+        # prepare extra kwargs for the scheduler step, since not all schedulers have the same signature
+        # eta (η) is only used with the DDIMScheduler, it will be ignored for other schedulers.
+        # eta corresponds to η in DDIM paper: https://arxiv.org/abs/2010.02502
+        # and should be between [0, 1]
+        accepts_eta = "eta" in set(inspect.signature(self.scheduler.step).parameters.keys())
+        extra_step_kwargs = {}
+        if accepts_eta:
+            extra_step_kwargs["eta"] = eta
+
+        timestep_dtype = next(
+            (input.type for input in self.unet.model.get_inputs() if input.name == "timestep"), "tensor(float)"
+        )
+        timestep_dtype = ORT_TO_NP_TYPE[timestep_dtype]
+
+        # panorama additions
+        views = self.get_views(height, width)
+        count = np.zeros_like(latents)
+        value = np.zeros_like(latents)
+
+        for i, t in enumerate(self.progress_bar(self.scheduler.timesteps)):
+            count.fill(0)
+            value.fill(0)
+
+            for h_start, h_end, w_start, w_end in views:
+                # get the latents corresponding to the current view coordinates
+                latents_for_view = latents[:, :, h_start:h_end, w_start:w_end]
+
+                # expand the latents if we are doing classifier free guidance
+                latent_model_input = np.concatenate([latents_for_view] * 2) if do_classifier_free_guidance else latents_for_view
+                latent_model_input = self.scheduler.scale_model_input(torch.from_numpy(latent_model_input), t)
+                latent_model_input = latent_model_input.cpu().numpy()
+
+                # predict the noise residual
+                timestep = np.array([t], dtype=timestep_dtype)
+                noise_pred = self.unet(sample=latent_model_input, timestep=timestep, encoder_hidden_states=prompt_embeds)
+                noise_pred = noise_pred[0]
+
+                # perform guidance
+                if do_classifier_free_guidance:
+                    noise_pred_uncond, noise_pred_text = np.split(noise_pred, 2)
+                    noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
+
+                # compute the previous noisy sample x_t -> x_t-1
+                scheduler_output = self.scheduler.step(
+                    torch.from_numpy(noise_pred), t, torch.from_numpy(latents_for_view), **extra_step_kwargs
+                )
+                latents_view_denoised = scheduler_output.prev_sample.numpy()
+
+                value[:, :, h_start:h_end, w_start:w_end] += latents_view_denoised
+                count[:, :, h_start:h_end, w_start:w_end] += 1
+
+            # take the MultiDiffusion step. Eq. 5 in MultiDiffusion paper: https://arxiv.org/abs/2302.08113
+            latents = np.where(count > 0, value / count, value)
+
+            # call the callback, if provided
+            if callback is not None and i % callback_steps == 0:
+                callback(i, t, latents)
+
+        latents = 1 / 0.18215 * latents
+        # image = self.vae_decoder(latent_sample=latents)[0]
+        # it seems likes there is a strange result for using half-precision vae decoder if batchsize>1
+        image = np.concatenate(
+            [self.vae_decoder(latent_sample=latents[i : i + 1])[0] for i in range(latents.shape[0])]
+        )
+
+        image = np.clip(image / 2 + 0.5, 0, 1)
+        image = image.transpose((0, 2, 3, 1))
+
+        if self.safety_checker is not None:
+            safety_checker_input = self.feature_extractor(
+                self.numpy_to_pil(image), return_tensors="np"
+            ).pixel_values.astype(image.dtype)
+
+            images, has_nsfw_concept = [], []
+            for i in range(image.shape[0]):
+                image_i, has_nsfw_concept_i = self.safety_checker(
+                    clip_input=safety_checker_input[i : i + 1], images=image[i : i + 1]
+                )
+                images.append(image_i)
+                has_nsfw_concept.append(has_nsfw_concept_i[0])
+            image = np.concatenate(images)
+        else:
+            has_nsfw_concept = None
+
+        if output_type == "pil":
+            image = self.numpy_to_pil(image)
+
+        if not return_dict:
+            return (image, has_nsfw_concept)
+
+        return StableDiffusionPipelineOutput(images=image, nsfw_content_detected=has_nsfw_concept)

api/onnx_web/diffusers/run.py

@@ -47,7 +47,9 @@ def run_loopback(
     loopback_progress = ChainProgress.from_progress(progress)
 
     # load img2img pipeline once
-    pipe_type = "lpw" if params.lpw() else "img2img"
+    pipe_type = params.get_valid_pipeline("img2img")
+    logger.debug("using %s pipeline for loopback", pipe_type)
+
     pipe = pipeline or load_pipeline(
         server,
         pipe_type,
@@ -59,8 +61,7 @@ def run_loopback(
     )
 
     def loopback_iteration(source: Image.Image):
-        if params.lpw():
+        if pipe_type in ["lpw", "panorama"]:
-            logger.debug("using LPW pipeline for loopback")
             rng = torch.manual_seed(params.seed)
             result = pipe.img2img(
                 source,
@@ -76,7 +77,6 @@ def run_loopback(
             )
             return result.images[0]
         else:
-            logger.debug("using img2img pipeline for loopback")
             rng = np.random.RandomState(params.seed)
             result = pipe(
                 params.prompt,
@@ -134,7 +134,9 @@ def run_highres(
         )
 
     # load img2img pipeline once
-    pipe_type = "lpw" if params.lpw() else "img2img"
+    pipe_type = params.get_valid_pipeline("img2img")
+    logger.debug("using %s pipeline for highres", pipe_type)
+
     highres_pipe = pipeline or load_pipeline(
         server,
         pipe_type,
@@ -172,8 +174,7 @@ def run_highres(
                 callback=highres_progress,
             )
 
-        if params.lpw():
+        if pipe_type in ["lpw", "panorama"]:
-            logger.debug("using LPW pipeline for highres")
             rng = torch.manual_seed(params.seed)
             result = highres_pipe.img2img(
                 tile,
@@ -189,7 +190,6 @@ def run_highres(
             )
             return result.images[0]
         else:
-            logger.debug("using img2img pipeline for highres")
             rng = np.random.RandomState(params.seed)
             result = highres_pipe(
                 params.prompt,
@@ -236,7 +236,9 @@ def run_txt2img_pipeline(
     latents = get_latents_from_seed(params.seed, size, batch=params.batch)
     prompt_pairs, loras, inversions = parse_prompt(params)
 
-    pipe_type = params.pipeline # TODO: allow txt2img, panorama, etc, but filter out the others
+    pipe_type = params.get_valid_pipeline("txt2img")
+    logger.debug("using %s pipeline for txt2img", pipe_type)
+
     pipe = load_pipeline(
         server,
         pipe_type,
@@ -248,8 +250,7 @@ def run_txt2img_pipeline(
     )
     progress = job.get_progress_callback()
 
-    if params.lpw():
+    if pipe_type in ["lpw", "panorama"]:
-        logger.debug("using LPW pipeline for txt2img")
         rng = torch.manual_seed(params.seed)
         result = pipe.text2img(
             params.prompt,
@@ -345,9 +346,10 @@ def run_img2img_pipeline(
             logger.debug("running source filter: %s", f.__name__)
             source = f(server, source)
 
+    pipe_type = params.get_valid_pipeline("img2img")
     pipe = load_pipeline(
         server,
-        params.pipeline,  # this is one of the only places this can actually vary between different pipelines
+        pipe_type,
         params.model,
         params.scheduler,
         job.get_device(),
@@ -357,15 +359,17 @@ def run_img2img_pipeline(
     )
 
     pipe_params = {}
-    if params.pipeline == "controlnet":
+    if pipe_type == "controlnet":
         pipe_params["controlnet_conditioning_scale"] = strength
-    elif params.pipeline == "img2img":
+    elif pipe_type == "img2img":
         pipe_params["strength"] = strength
-    elif params.pipeline == "pix2pix":
+    elif pipe_type == "panorama":
+        pipe_params["strength"] = strength
+    elif pipe_type == "pix2pix":
         pipe_params["image_guidance_scale"] = strength
 
     progress = job.get_progress_callback()
-    if params.lpw():
+    if pipe_type in ["lpw", "panorama"]:
         logger.debug("using LPW pipeline for img2img")
         rng = torch.manual_seed(params.seed)
         result = pipe.img2img(

api/onnx_web/params.py

@@ -220,6 +220,27 @@ class ImageParams:
     def do_cfg(self):
         return self.cfg > 1.0
 
+    def get_valid_pipeline(self, group: str, pipeline: str = None) -> str:
+        pipeline = pipeline or self.pipeline
+
+        # if the correct pipeline was already requested, simply use that
+        if group == pipeline:
+            return pipeline
+
+        # otherwise, check for additional allowed pipelines
+        if group == "img2img":
+            if pipeline in ["controlnet", "lpw", "panorama", "pix2pix"]:
+                return pipeline
+        elif group == "inpaint":
+            if pipeline in ["controlnet"]:
+                return pipeline
+        elif group == "txt2img":
+            if pipeline in ["panorama"]:
+                return pipeline
+
+        logger.debug("pipeline %s is not valid for %s", pipeline, group)
+        return group
+
     def lpw(self):
         return self.pipeline == "lpw"