[Img2Img2] Re-add K LMS scheduler (#340)

src/diffusers/pipelines/stable_diffusion/pipeline_stable_diffusion_img2img.py

@@ -5,12 +5,11 @@ import numpy as np
 import torch
 
 import PIL
-from tqdm.auto import tqdm
 from transformers import CLIPFeatureExtractor, CLIPTextModel, CLIPTokenizer
 
 from ...models import AutoencoderKL, UNet2DConditionModel
 from ...pipeline_utils import DiffusionPipeline
-from ...schedulers import DDIMScheduler, PNDMScheduler
+from ...schedulers import DDIMScheduler, LMSDiscreteScheduler, PNDMScheduler
 from .safety_checker import StableDiffusionSafetyChecker
 
 
@@ -31,7 +30,7 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         text_encoder: CLIPTextModel,
         tokenizer: CLIPTokenizer,
         unet: UNet2DConditionModel,
-        scheduler: Union[DDIMScheduler, PNDMScheduler],
+        scheduler: Union[DDIMScheduler, PNDMScheduler, LMSDiscreteScheduler],
         safety_checker: StableDiffusionSafetyChecker,
         feature_extractor: CLIPFeatureExtractor,
     ):
@@ -93,12 +92,17 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
         # get the original timestep using init_timestep
         init_timestep = int(num_inference_steps * strength) + offset
         init_timestep = min(init_timestep, num_inference_steps)
-        timesteps = self.scheduler.timesteps[-init_timestep]
-        timesteps = torch.tensor([timesteps] * batch_size, dtype=torch.long, device=self.device)
+        if isinstance(self.scheduler, LMSDiscreteScheduler):
+            timesteps = torch.tensor(
+                [num_inference_steps - init_timestep] * batch_size, dtype=torch.long, device=self.device
+            )
+        else:
+            timesteps = self.scheduler.timesteps[-init_timestep]
+            timesteps = torch.tensor([timesteps] * batch_size, dtype=torch.long, device=self.device)
 
         # add noise to latents using the timesteps
         noise = torch.randn(init_latents.shape, generator=generator, device=self.device)
-        init_latents = self.scheduler.add_noise(init_latents, noise, timesteps)
+        init_latents = self.scheduler.add_noise(init_latents, noise, timesteps).to(self.device)
 
         # get prompt text embeddings
         text_input = self.tokenizer(
@@ -137,11 +141,22 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
             extra_step_kwargs["eta"] = eta
 
         latents = init_latents
+
         t_start = max(num_inference_steps - init_timestep + offset, 0)
-        for i, t in tqdm(enumerate(self.scheduler.timesteps[t_start:])):
+        for i, t in enumerate(self.progress_bar(self.scheduler.timesteps[t_start:])):
+            t_index = t_start + i
+
             # expand the latents if we are doing classifier free guidance
             latent_model_input = torch.cat([latents] * 2) if do_classifier_free_guidance else latents
 
+            # if we use LMSDiscreteScheduler, let's make sure latents are mulitplied by sigmas
+            if isinstance(self.scheduler, LMSDiscreteScheduler):
+                sigma = self.scheduler.sigmas[t_index]
+                # the model input needs to be scaled to match the continuous ODE formulation in K-LMS
+                latent_model_input = latent_model_input / ((sigma**2 + 1) ** 0.5)
+                latent_model_input = latent_model_input.to(self.unet.dtype)
+                t = t.to(self.unet.dtype)
+
             # predict the noise residual
             noise_pred = self.unet(latent_model_input, t, encoder_hidden_states=text_embeddings)["sample"]
 
@@ -151,11 +166,14 @@ class StableDiffusionImg2ImgPipeline(DiffusionPipeline):
                 noise_pred = noise_pred_uncond + guidance_scale * (noise_pred_text - noise_pred_uncond)
 
             # compute the previous noisy sample x_t -> x_t-1
-            latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs)["prev_sample"]
+            if isinstance(self.scheduler, LMSDiscreteScheduler):
+                latents = self.scheduler.step(noise_pred, t_index, latents, **extra_step_kwargs)["prev_sample"]
+            else:
+                latents = self.scheduler.step(noise_pred, t, latents, **extra_step_kwargs)["prev_sample"]
 
         # scale and decode the image latents with vae
         latents = 1 / 0.18215 * latents
-        image = self.vae.decode(latents)
+        image = self.vae.decode(latents.to(self.vae.dtype))
 
         image = (image / 2 + 0.5).clamp(0, 1)
         image = image.cpu().permute(0, 2, 3, 1).numpy()

tests/test_pipelines.py

@@ -445,6 +445,49 @@ class PipelineFastTests(unittest.TestCase):
         expected_slice = np.array([0.4492, 0.3865, 0.4222, 0.5854, 0.5139, 0.4379, 0.4193, 0.48, 0.4218])
         assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-2
 
+    def test_stable_diffusion_img2img_k_lms(self):
+        device = "cpu"  # ensure determinism for the device-dependent torch.Generator
+        unet = self.dummy_cond_unet
+        scheduler = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear")
+
+        vae = self.dummy_vae
+        bert = self.dummy_text_encoder
+        tokenizer = CLIPTokenizer.from_pretrained("hf-internal-testing/tiny-random-clip")
+
+        init_image = self.dummy_image.to(device)
+
+        # make sure here that pndm scheduler skips prk
+        sd_pipe = StableDiffusionImg2ImgPipeline(
+            unet=unet,
+            scheduler=scheduler,
+            vae=vae,
+            text_encoder=bert,
+            tokenizer=tokenizer,
+            safety_checker=self.dummy_safety_checker,
+            feature_extractor=self.dummy_extractor,
+        )
+        sd_pipe = sd_pipe.to(device)
+        sd_pipe.set_progress_bar_config(disable=None)
+
+        prompt = "A painting of a squirrel eating a burger"
+        generator = torch.Generator(device=device).manual_seed(0)
+        output = sd_pipe(
+            [prompt],
+            generator=generator,
+            guidance_scale=6.0,
+            num_inference_steps=2,
+            output_type="np",
+            init_image=init_image,
+        )
+
+        image = output["sample"]
+
+        image_slice = image[0, -3:, -3:, -1]
+
+        assert image.shape == (1, 32, 32, 3)
+        expected_slice = np.array([0.4367, 0.4986, 0.4372, 0.6706, 0.5665, 0.444, 0.5864, 0.6019, 0.5203])
+        assert np.abs(image_slice.flatten() - expected_slice).max() < 1e-2
+
     def test_stable_diffusion_inpaint(self):
         device = "cpu"  # ensure determinism for the device-dependent torch.Generator
         unet = self.dummy_cond_unet
@@ -892,7 +935,7 @@ class PipelineTesterMixin(unittest.TestCase):
     def test_stable_diffusion_img2img_pipeline(self):
         ds = load_dataset("hf-internal-testing/diffusers-images", split="train")
 
-        init_image = ds[1]["image"].resize((768, 512))
+        init_image = ds[2]["image"].resize((768, 512))
         output_image = ds[0]["image"].resize((768, 512))
 
         model_id = "CompVis/stable-diffusion-v1-4"
@@ -915,12 +958,40 @@ class PipelineTesterMixin(unittest.TestCase):
 
     @slow
     @unittest.skipIf(torch_device == "cpu", "Stable diffusion is supposed to run on GPU")
-    def test_stable_diffusion_in_paint_pipeline(self):
+    def test_stable_diffusion_img2img_pipeline_k_lms(self):
         ds = load_dataset("hf-internal-testing/diffusers-images", split="train")
 
         init_image = ds[2]["image"].resize((768, 512))
-        mask_image = ds[3]["image"].resize((768, 512))
-        output_image = ds[4]["image"].resize((768, 512))
+        output_image = ds[1]["image"].resize((768, 512))
+
+        lms = LMSDiscreteScheduler(beta_start=0.00085, beta_end=0.012, beta_schedule="scaled_linear")
+
+        model_id = "CompVis/stable-diffusion-v1-4"
+        pipe = StableDiffusionImg2ImgPipeline.from_pretrained(model_id, scheduler=lms, use_auth_token=True)
+        pipe.to(torch_device)
+        pipe.set_progress_bar_config(disable=None)
+
+        prompt = "A fantasy landscape, trending on artstation"
+
+        generator = torch.Generator(device=torch_device).manual_seed(0)
+        image = pipe(prompt=prompt, init_image=init_image, strength=0.75, guidance_scale=7.5, generator=generator)[
+            "sample"
+        ][0]
+
+        expected_array = np.array(output_image)
+        sampled_array = np.array(image)
+
+        assert sampled_array.shape == (512, 768, 3)
+        assert np.max(np.abs(sampled_array - expected_array)) < 1e-4
+
+    @slow
+    @unittest.skipIf(torch_device == "cpu", "Stable diffusion is supposed to run on GPU")
+    def test_stable_diffusion_in_paint_pipeline(self):
+        ds = load_dataset("hf-internal-testing/diffusers-images", split="train")
+
+        init_image = ds[3]["image"].resize((768, 512))
+        mask_image = ds[4]["image"].resize((768, 512))
+        output_image = ds[5]["image"].resize((768, 512))
 
         model_id = "CompVis/stable-diffusion-v1-4"
         pipe = StableDiffusionInpaintPipeline.from_pretrained(model_id, use_auth_token=True)