Typo fixed

Author: Werner Oswald

ldm/models/diffusion/ddim.py

@@ -150,6 +150,8 @@ def ddim_sampling(self, cond, shape,
                 assert x0 is not None
                 img_orig = self.model.q_sample(x0, ts)  # TODO: deterministic forward pass?
                 img = img_orig * mask + (1. - mask) * img
+                #tmp_mask = (mask > (1 - (step / 1000))) * 1
+                #img = img_orig_with_noise * tmp_mask + (1. - tmp_mask) * img
 
             outs = self.p_sample_ddim(img, cond, ts, index=index, use_original_steps=ddim_use_original_steps,
                                       quantize_denoised=quantize_denoised, temperature=temperature,
@@ -159,6 +161,7 @@ def ddim_sampling(self, cond, shape,
                                       unconditional_conditioning=unconditional_conditioning,
                                       dynamic_threshold=dynamic_threshold)
             img, pred_x0 = outs
+            #if callback: callback(i)
             if callback:
                 img = callback(i, img, pred_x0)
             if img_callback: img_callback(pred_x0, i)
@@ -292,7 +295,7 @@ def stochastic_encode(self, x0, t, use_original_steps=False, noise=None):
 
     @torch.no_grad()
     def decode(self, x_latent, cond, t_start, unconditional_guidance_scale=1.0, unconditional_conditioning=None,
-               use_original_steps=False):
+               use_original_steps=False, img_callback=None):
 
         timesteps = np.arange(self.ddpm_num_timesteps) if use_original_steps else self.ddim_timesteps
         timesteps = timesteps[:t_start]
@@ -309,4 +312,7 @@ def decode(self, x_latent, cond, t_start, unconditional_guidance_scale=1.0, unco
             x_dec, _ = self.p_sample_ddim(x_dec, cond, ts, index=index, use_original_steps=use_original_steps,
                                           unconditional_guidance_scale=unconditional_guidance_scale,
                                           unconditional_conditioning=unconditional_conditioning)
+
+            if img_callback: img_callback(x_dec, i)
+
         return x_dec

requirements.txt

@@ -1,4 +0,0 @@
-yaml~=0.2.5
-tqdm~=4.63.0
-setuptools~=61.2.0
-requests~=2.27.1

scripts/tools/config/webui_streamlit.yaml

@@ -6,6 +6,8 @@ general:
   gpu: 0
   outdir: outputs
   ckpt: "content/models/sd-v1-4.ckpt"
+  clip: "content/models/sd-clip-vit-l14-img-embed_ema_only.ckpt"
+  finetune: "configs/stable-diffusion/sd-image-condition-finetune.yaml"
   default_model: "Stable Diffusion v1.4"
   default_model_config: "configs/stable-diffusion/v1-inference.yaml"
   default_model_path: "content/models/sd-v1-4.ckpt"

scripts/tools/modelloader.py

@@ -86,6 +86,7 @@ def load_model_from_config(config, ckpt, verbose=False):
 	model.eval()
 	return model
 
+#@st.cache
 def load_models(continue_prev_run=False, use_GFPGAN=False, use_RealESRGAN=False, RealESRGAN_model="RealESRGAN_x4plus"):
 	"""Load the different models. We also reuse the models that are already in memory to speed things up instead of loading them again. """
 

scripts/tools/node_func.py

@@ -65,8 +65,8 @@ def load_var_model_from_config(config_var, ckpt_var, device, verbose=False, half
 
 def variations(input_im, outdir, var_samples, var_plms, v_cfg_scale, v_steps, v_W, v_H, v_ddim_eta, v_GFPGAN, v_bg_upsampling, v_upscale):
     #im_path="data/example_conditioning/superresolution/sample_0.jpg",
-    ckpt_var="/gdrive/MyDrive/sd-clip-vit-l14-img-embed_ema_only.ckpt"
-    config_var="/content/deforum-sd-ui-colab/configs/stable-diffusion/sd-image-condition-finetune.yaml"
+    ckpt_var=st.session_state.defaults.general.clip
+    config_var=st.session_state.defaults.general.finetune
     outpath=outdir
     scale=v_cfg_scale
     h=v_H

scripts/ui/img2img.py

@@ -275,16 +275,17 @@ def layoutFunc():
 		st.session_state[st.session_state['generation_mode']]['prompt_matrix'] = None
 
 		if uploaded_images:
-			image = Image.open(uploaded_images).convert('RGBA')
+			image = Image.open(uploaded_images).convert('RGB')
 			st.session_state["img2img"]["new_img"] = image.resize((st.session_state["img2img"]["width"], st.session_state["img2img"]["height"]))
 			# img_array = np.array(image) # if you want to pass it to OpenCV
 			new_mask = None
 			if st.session_state["img2img"]["uploaded_masks"]:
-				mask = Image.open(st.session_state["img2img"]["uploaded_masks"]).convert('RGBA')
+				mask = Image.open(st.session_state["img2img"]["uploaded_masks"]).convert('RGB')
 				new_mask = mask.resize((st.session_state["img2img"]["width"], st.session_state["img2img"]["height"]))
 
 			try:
-
+				output_imgs = variations(image, outdir='output', var_samples=4, var_plms='k_lms', v_cfg_scale=7.5, v_steps=20, v_W=512, v_H=512, v_ddim_eta=0, v_GFPGAN=False, v_bg_upsampling=False, v_upscale=1)
+				"""
 				output_images, seed, info, stats = img2img(prompt=st.session_state["img2img"]["prompt"],
 														   init_info=st.session_state["img2img"]["new_img"],
 														   init_info_mask=new_mask,
@@ -316,10 +317,10 @@ def layoutFunc():
 														   use_RealESRGAN=st.session_state["img2img"]["use_RealESRGAN"] if not st.session_state["img2img"]["loopback"] else False,
 														   loopback=st.session_state["img2img"]["loopback"]
 														   )
-
+										
 				# show a message when the generation is complete.
 				message.success('Render Complete: ' + info + '; Stats: ' + stats, icon="✅")
-
+				"""
 			except (StopException, KeyError) as e:
 				print(e)
 				print(f"Received Streamlit StopException")

setup_env.cmd

@@ -70,6 +70,6 @@ IF EXIST "content\models\dpt_large-midas-2f21e586.pt" (
   echo all ready to fly
 
 ) ELSE (
-  echo Your model file does not exist! Place it in 'content\models\dpt_large-midas-2f21e586.pt' with the name 'model.ckpt'.
+  echo Your model file does not exist! Place it in 'content\models\dpt_large-midas-2f21e586.pt' with the name 'dpt_large-midas-2f21e586.pt'.
   pause
 )