Merge pull request #14390 from wangqyqq/sdxl-inpaint

Supporting for SDXL-Inpaint Model

modules/processing.py

@@ -113,6 +113,21 @@ def txt2img_image_conditioning(sd_model, x, width, height):
         return x.new_zeros(x.shape[0], 2*sd_model.noise_augmentor.time_embed.dim, dtype=x.dtype, device=x.device)
 
     else:
+        sd = sd_model.model.state_dict()
+        diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+        if diffusion_model_input is not None:
+            if diffusion_model_input.shape[1] == 9:
+                # The "masked-image" in this case will just be all 0.5 since the entire image is masked.
+                image_conditioning = torch.ones(x.shape[0], 3, height, width, device=x.device) * 0.5
+                image_conditioning = images_tensor_to_samples(image_conditioning,
+                                                              approximation_indexes.get(opts.sd_vae_encode_method))
+
+                # Add the fake full 1s mask to the first dimension.
+                image_conditioning = torch.nn.functional.pad(image_conditioning, (0, 0, 0, 0, 1, 0), value=1.0)
+                image_conditioning = image_conditioning.to(x.dtype)
+
+                return image_conditioning
+
         # Dummy zero conditioning if we're not using inpainting or unclip models.
         # Still takes up a bit of memory, but no encoder call.
         # Pretty sure we can just make this a 1x1 image since its not going to be used besides its batch size.
@@ -371,6 +386,12 @@ class StableDiffusionProcessing:
         if self.sampler.conditioning_key == "crossattn-adm":
             return self.unclip_image_conditioning(source_image)
 
+        sd = self.sampler.model_wrap.inner_model.model.state_dict()
+        diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+        if diffusion_model_input is not None:
+            if diffusion_model_input.shape[1] == 9:
+                return self.inpainting_image_conditioning(source_image, latent_image, image_mask=image_mask)
+
         # Dummy zero conditioning if we're not using inpainting or depth model.
         return latent_image.new_zeros(latent_image.shape[0], 5, 1, 1)
 

modules/sd_models_config.py

@@ -15,6 +15,7 @@ config_sd2v = os.path.join(sd_repo_configs_path, "v2-inference-v.yaml")
 config_sd2_inpainting = os.path.join(sd_repo_configs_path, "v2-inpainting-inference.yaml")
 config_sdxl = os.path.join(sd_xl_repo_configs_path, "sd_xl_base.yaml")
 config_sdxl_refiner = os.path.join(sd_xl_repo_configs_path, "sd_xl_refiner.yaml")
+config_sdxl_inpainting = os.path.join(sd_configs_path, "sd_xl_inpaint.yaml")
 config_depth_model = os.path.join(sd_repo_configs_path, "v2-midas-inference.yaml")
 config_unclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-l-inference.yaml")
 config_unopenclip = os.path.join(sd_repo_configs_path, "v2-1-stable-unclip-h-inference.yaml")
@@ -71,7 +72,10 @@ def guess_model_config_from_state_dict(sd, filename):
     sd2_variations_weight = sd.get('embedder.model.ln_final.weight', None)
 
     if sd.get('conditioner.embedders.1.model.ln_final.weight', None) is not None:
-        return config_sdxl
+        if diffusion_model_input.shape[1] == 9:
+            return config_sdxl_inpainting
+        else:
+            return config_sdxl
     if sd.get('conditioner.embedders.0.model.ln_final.weight', None) is not None:
         return config_sdxl_refiner
     elif sd.get('depth_model.model.pretrained.act_postprocess3.0.project.0.bias', None) is not None:

modules/sd_models_xl.py

@@ -34,6 +34,12 @@ def get_learned_conditioning(self: sgm.models.diffusion.DiffusionEngine, batch:
 
 
 def apply_model(self: sgm.models.diffusion.DiffusionEngine, x, t, cond):
+    sd = self.model.state_dict()
+    diffusion_model_input = sd.get('diffusion_model.input_blocks.0.0.weight', None)
+    if diffusion_model_input is not None:
+        if diffusion_model_input.shape[1] == 9:
+            x = torch.cat([x] + cond['c_concat'], dim=1)
+
     return self.model(x, t, cond)