diff --git a/.ci/update_windows/update.py b/.ci/update_windows/update.py
index ef9374c441d..127247b2fc0 100755
--- a/.ci/update_windows/update.py
+++ b/.ci/update_windows/update.py
@@ -1,6 +1,9 @@
 import pygit2
 from datetime import datetime
 import sys
+import os
+import shutil
+import filecmp
 
 def pull(repo, remote_name='origin', branch='master'):
     for remote in repo.remotes:
@@ -42,7 +45,8 @@ def pull(repo, remote_name='origin', branch='master'):
                 raise AssertionError('Unknown merge analysis result')
 
 pygit2.option(pygit2.GIT_OPT_SET_OWNER_VALIDATION, 0)
-repo = pygit2.Repository(str(sys.argv[1]))
+repo_path = str(sys.argv[1])
+repo = pygit2.Repository(repo_path)
 ident = pygit2.Signature('comfyui', 'comfy@ui')
 try:
     print("stashing current changes")
@@ -51,7 +55,10 @@ def pull(repo, remote_name='origin', branch='master'):
     print("nothing to stash")
 backup_branch_name = 'backup_branch_{}'.format(datetime.today().strftime('%Y-%m-%d_%H_%M_%S'))
 print("creating backup branch: {}".format(backup_branch_name))
-repo.branches.local.create(backup_branch_name, repo.head.peel())
+try:
+    repo.branches.local.create(backup_branch_name, repo.head.peel())
+except:
+    pass
 
 print("checking out master branch")
 branch = repo.lookup_branch('master')
@@ -63,3 +70,41 @@ def pull(repo, remote_name='origin', branch='master'):
 
 print("Done!")
 
+self_update = True
+if len(sys.argv) > 2:
+    self_update = '--skip_self_update' not in sys.argv
+
+update_py_path = os.path.realpath(__file__)
+repo_update_py_path = os.path.join(repo_path, ".ci/update_windows/update.py")
+
+cur_path = os.path.dirname(update_py_path)
+
+
+req_path = os.path.join(cur_path, "current_requirements.txt")
+repo_req_path = os.path.join(repo_path, "requirements.txt")
+
+
+def files_equal(file1, file2):
+    try:
+        return filecmp.cmp(file1, file2, shallow=False)
+    except:
+        return False
+
+def file_size(f):
+    try:
+        return os.path.getsize(f)
+    except:
+        return 0
+
+
+if self_update and not files_equal(update_py_path, repo_update_py_path) and file_size(repo_update_py_path) > 10:
+    shutil.copy(repo_update_py_path, os.path.join(cur_path, "update_new.py"))
+    exit()
+
+if not os.path.exists(req_path) or not files_equal(repo_req_path, req_path):
+    import subprocess
+    try:
+        subprocess.check_call([sys.executable, '-s', '-m', 'pip', 'install', '-r', repo_req_path])
+        shutil.copy(repo_req_path, req_path)
+    except:
+        pass
diff --git a/.ci/update_windows/update_comfyui.bat b/.ci/update_windows/update_comfyui.bat
index 60d1e694fa4..bb08c0de0c7 100755
--- a/.ci/update_windows/update_comfyui.bat
+++ b/.ci/update_windows/update_comfyui.bat
@@ -1,2 +1,8 @@
+@echo off
 ..\python_embeded\python.exe .\update.py ..\ComfyUI\
-pause
+if exist update_new.py (
+  move /y update_new.py update.py
+  echo Running updater again since it got updated.
+  ..\python_embeded\python.exe .\update.py ..\ComfyUI\ --skip_self_update
+)
+if "%~1"=="" pause
diff --git a/.ci/update_windows/update_comfyui_and_python_dependencies.bat b/.ci/update_windows/update_comfyui_and_python_dependencies.bat
deleted file mode 100755
index b7308550d1d..00000000000
--- a/.ci/update_windows/update_comfyui_and_python_dependencies.bat
+++ /dev/null
@@ -1,3 +0,0 @@
-..\python_embeded\python.exe .\update.py ..\ComfyUI\
-..\python_embeded\python.exe -s -m pip install --upgrade torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu117 xformers -r ../ComfyUI/requirements.txt pygit2
-pause
diff --git a/.ci/update_windows_cu118/update_comfyui_and_python_dependencies.bat b/.ci/update_windows_cu118/update_comfyui_and_python_dependencies.bat
deleted file mode 100755
index c33adc0a7b8..00000000000
--- a/.ci/update_windows_cu118/update_comfyui_and_python_dependencies.bat
+++ /dev/null
@@ -1,11 +0,0 @@
-@echo off
-..\python_embeded\python.exe .\update.py ..\ComfyUI\
-echo
-echo This will try to update pytorch and all python dependencies, if you get an error wait for pytorch/xformers to fix their stuff
-echo You should not be running this anyways unless you really have to
-echo
-echo If you just want to update normally, close this and run update_comfyui.bat instead.
-echo
-pause
-..\python_embeded\python.exe -s -m pip install --upgrade torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118 xformers -r ../ComfyUI/requirements.txt pygit2
-pause
diff --git a/.github/workflows/windows_release_cu118_dependencies.yml b/.github/workflows/windows_release_cu118_dependencies.yml
deleted file mode 100644
index 75c42b624a9..00000000000
--- a/.github/workflows/windows_release_cu118_dependencies.yml
+++ /dev/null
@@ -1,71 +0,0 @@
-name: "Windows Release cu118 dependencies"
-
-on:
-  workflow_dispatch:
-#  push:
-#    branches:
-#      - master
-
-jobs:
-  build_dependencies:
-    env:
-        # you need at least cuda 5.0 for some of the stuff compiled here.
-        TORCH_CUDA_ARCH_LIST: "5.0+PTX 6.0 6.1 7.0 7.5 8.0 8.6 8.9"
-        FORCE_CUDA: 1
-        MAX_JOBS: 1 # will crash otherwise
-        DISTUTILS_USE_SDK: 1 # otherwise distutils will complain on windows about multiple versions of msvc
-        XFORMERS_BUILD_TYPE: "Release"
-    runs-on: windows-latest
-    steps:
-        - name: Cache Built Dependencies
-          uses: actions/cache@v3
-          id: cache-cu118_python_stuff
-          with:
-            path: cu118_python_deps.tar
-            key: ${{ runner.os }}-build-cu118
-
-        - if: steps.cache-cu118_python_stuff.outputs.cache-hit != 'true'
-          uses: actions/checkout@v3
-
-        - if: steps.cache-cu118_python_stuff.outputs.cache-hit != 'true'
-          uses: actions/setup-python@v4
-          with:
-            python-version: '3.10.9'
-
-        - if: steps.cache-cu118_python_stuff.outputs.cache-hit != 'true'
-          uses: comfyanonymous/cuda-toolkit@test
-          id: cuda-toolkit
-          with:
-            cuda: '11.8.0'
-        # copied from xformers github
-        - name: Setup MSVC
-          uses: ilammy/msvc-dev-cmd@v1
-        - name: Configure Pagefile
-          # windows runners will OOM with many CUDA architectures
-          # we cheat here with a page file
-          uses: al-cheb/configure-pagefile-action@v1.3
-          with:
-            minimum-size: 2GB
-        # really unfortunate: https://github.com/ilammy/msvc-dev-cmd#name-conflicts-with-shell-bash
-        - name: Remove link.exe
-          shell: bash
-          run: rm /usr/bin/link
-
-        - if: steps.cache-cu118_python_stuff.outputs.cache-hit != 'true'
-          shell: bash
-          run: |
-            python -m pip wheel --no-cache-dir torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118 -r requirements.txt pygit2 -w ./temp_wheel_dir
-            python -m pip install --no-cache-dir ./temp_wheel_dir/*
-            echo installed basic
-            git clone --recurse-submodules https://github.com/facebookresearch/xformers.git
-            cd xformers
-            python -m pip install --no-cache-dir wheel setuptools twine
-            echo building xformers
-            python setup.py bdist_wheel -d ../temp_wheel_dir/
-            cd ..
-            rm -rf xformers
-            ls -lah temp_wheel_dir
-            mv temp_wheel_dir cu118_python_deps
-            tar cf cu118_python_deps.tar cu118_python_deps
-
-
diff --git a/.github/workflows/windows_release_cu118_dependencies_2.yml b/.github/workflows/windows_release_cu118_dependencies_2.yml
deleted file mode 100644
index a7760b21e15..00000000000
--- a/.github/workflows/windows_release_cu118_dependencies_2.yml
+++ /dev/null
@@ -1,37 +0,0 @@
-name: "Windows Release cu118 dependencies 2"
-
-on:
-  workflow_dispatch:
-    inputs:
-      xformers:
-        description: 'xformers version'
-        required: true
-        type: string
-        default: "xformers"
-
-#  push:
-#    branches:
-#      - master
-
-jobs:
-  build_dependencies:
-    runs-on: windows-latest
-    steps:
-        - uses: actions/checkout@v3
-        - uses: actions/setup-python@v4
-          with:
-            python-version: '3.10.9'
-
-        - shell: bash
-          run: |
-            python -m pip wheel --no-cache-dir torch torchvision torchaudio ${{ inputs.xformers }} --extra-index-url https://download.pytorch.org/whl/cu118 -r requirements.txt pygit2 -w ./temp_wheel_dir
-            python -m pip install --no-cache-dir ./temp_wheel_dir/*
-            echo installed basic
-            ls -lah temp_wheel_dir
-            mv temp_wheel_dir cu118_python_deps
-            tar cf cu118_python_deps.tar cu118_python_deps
-
-        - uses: actions/cache/save@v3
-          with:
-            path: cu118_python_deps.tar
-            key: ${{ runner.os }}-build-cu118
diff --git a/.github/workflows/windows_release_cu118_package.yml b/.github/workflows/windows_release_cu118_package.yml
deleted file mode 100644
index 0f0fbf28039..00000000000
--- a/.github/workflows/windows_release_cu118_package.yml
+++ /dev/null
@@ -1,79 +0,0 @@
-name: "Windows Release cu118 packaging"
-
-on:
-  workflow_dispatch:
-#  push:
-#    branches:
-#      - master
-
-jobs:
-  package_comfyui:
-    permissions:
-        contents: "write"
-        packages: "write"
-        pull-requests: "read"
-    runs-on: windows-latest
-    steps:
-        - uses: actions/cache/restore@v3
-          id: cache
-          with:
-            path: cu118_python_deps.tar
-            key: ${{ runner.os }}-build-cu118
-        - shell: bash
-          run: |
-            mv cu118_python_deps.tar ../
-            cd ..
-            tar xf cu118_python_deps.tar
-            pwd
-            ls
-
-        - uses: actions/checkout@v3
-          with:
-            fetch-depth: 0
-            persist-credentials: false
-        - shell: bash
-          run: |
-            cd ..
-            cp -r ComfyUI ComfyUI_copy
-            curl https://www.python.org/ftp/python/3.10.9/python-3.10.9-embed-amd64.zip -o python_embeded.zip
-            unzip python_embeded.zip -d python_embeded
-            cd python_embeded
-            echo 'import site' >> ./python310._pth
-            curl https://bootstrap.pypa.io/get-pip.py -o get-pip.py
-            ./python.exe get-pip.py
-            ./python.exe -s -m pip install ../cu118_python_deps/*
-            sed -i '1i../ComfyUI' ./python310._pth
-            cd ..
-
-            git clone https://github.com/comfyanonymous/taesd
-            cp taesd/*.pth ./ComfyUI_copy/models/vae_approx/
-
-            mkdir ComfyUI_windows_portable
-            mv python_embeded ComfyUI_windows_portable
-            mv ComfyUI_copy ComfyUI_windows_portable/ComfyUI
-
-            cd ComfyUI_windows_portable
-
-            mkdir update
-            cp -r ComfyUI/.ci/update_windows/* ./update/
-            cp -r ComfyUI/.ci/update_windows_cu118/* ./update/
-            cp -r ComfyUI/.ci/windows_base_files/* ./
-
-            cd ..
-
-            "C:\Program Files\7-Zip\7z.exe" a -t7z -m0=lzma -mx=8 -mfb=64 -md=32m -ms=on -mf=BCJ2 ComfyUI_windows_portable.7z ComfyUI_windows_portable
-            mv ComfyUI_windows_portable.7z ComfyUI/new_ComfyUI_windows_portable_nvidia_cu118_or_cpu.7z
-
-            cd ComfyUI_windows_portable
-            python_embeded/python.exe -s ComfyUI/main.py --quick-test-for-ci --cpu
-
-            ls
-
-        - name: Upload binaries to release
-          uses: svenstaro/upload-release-action@v2
-          with:
-                repo_token: ${{ secrets.GITHUB_TOKEN }}
-                file: new_ComfyUI_windows_portable_nvidia_cu118_or_cpu.7z
-                tag: "latest"
-                overwrite: true
-
diff --git a/.github/workflows/windows_release_dependencies.yml b/.github/workflows/windows_release_dependencies.yml
index aafe8a21444..e0841fdf6c3 100644
--- a/.github/workflows/windows_release_dependencies.yml
+++ b/.github/workflows/windows_release_dependencies.yml
@@ -41,10 +41,9 @@ jobs:
         - shell: bash
           run: |
             echo "@echo off
-            ..\python_embeded\python.exe .\update.py ..\ComfyUI\\
+            call update_comfyui.bat nopause
             echo -
-            echo This will try to update pytorch and all python dependencies, if you get an error wait for pytorch/xformers to fix their stuff
-            echo You should not be running this anyways unless you really have to
+            echo This will try to update pytorch and all python dependencies.
             echo -
             echo If you just want to update normally, close this and run update_comfyui.bat instead.
             echo -
diff --git a/.github/workflows/windows_release_nightly_pytorch.yml b/.github/workflows/windows_release_nightly_pytorch.yml
index 90e09d27a53..56830935685 100644
--- a/.github/workflows/windows_release_nightly_pytorch.yml
+++ b/.github/workflows/windows_release_nightly_pytorch.yml
@@ -68,7 +68,7 @@ jobs:
             cp -r ComfyUI/.ci/update_windows/* ./update/
             cp -r ComfyUI/.ci/windows_base_files/* ./
 
-            echo "..\python_embeded\python.exe .\update.py ..\ComfyUI\\
+            echo "call update_comfyui.bat nopause
             ..\python_embeded\python.exe -s -m pip install --upgrade --pre torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cu${{ inputs.cu }} -r ../ComfyUI/requirements.txt pygit2
             pause" > ./update/update_comfyui_and_python_dependencies.bat
             cd ..
diff --git a/comfy/clip_model.py b/comfy/clip_model.py
index 9b82a246b2c..14f43c5687c 100644
--- a/comfy/clip_model.py
+++ b/comfy/clip_model.py
@@ -119,6 +119,9 @@ def __init__(self, config_dict, dtype, device, operations):
         super().__init__()
         self.num_layers = config_dict["num_hidden_layers"]
         self.text_model = CLIPTextModel_(config_dict, dtype, device, operations)
+        embed_dim = config_dict["hidden_size"]
+        self.text_projection = operations.Linear(embed_dim, embed_dim, bias=False, dtype=dtype, device=device)
+        self.text_projection.weight.copy_(torch.eye(embed_dim))
         self.dtype = dtype
 
     def get_input_embeddings(self):
@@ -128,7 +131,10 @@ def set_input_embeddings(self, embeddings):
         self.text_model.embeddings.token_embedding = embeddings
 
     def forward(self, *args, **kwargs):
-        return self.text_model(*args, **kwargs)
+        x = self.text_model(*args, **kwargs)
+        out = self.text_projection(x[2])
+        return (x[0], x[1], out, x[2])
+
 
 class CLIPVisionEmbeddings(torch.nn.Module):
     def __init__(self, embed_dim, num_channels=3, patch_size=14, image_size=224, dtype=None, device=None, operations=None):
diff --git a/comfy/controlnet.py b/comfy/controlnet.py
index 416197586a1..f859a50d4c3 100644
--- a/comfy/controlnet.py
+++ b/comfy/controlnet.py
@@ -287,13 +287,13 @@ class control_lora_ops(ControlLoraOps, comfy.ops.manual_cast):
         for k in sd:
             weight = sd[k]
             try:
-                comfy.utils.set_attr(self.control_model, k, weight)
+                comfy.utils.set_attr_param(self.control_model, k, weight)
             except:
                 pass
 
         for k in self.control_weights:
             if k not in {"lora_controlnet"}:
-                comfy.utils.set_attr(self.control_model, k, self.control_weights[k].to(dtype).to(comfy.model_management.get_torch_device()))
+                comfy.utils.set_attr_param(self.control_model, k, self.control_weights[k].to(dtype).to(comfy.model_management.get_torch_device()))
 
     def copy(self):
         c = ControlLora(self.control_weights, global_average_pooling=self.global_average_pooling)
diff --git a/comfy/diffusers_convert.py b/comfy/diffusers_convert.py
index a9eb9302f14..eb561933aaa 100644
--- a/comfy/diffusers_convert.py
+++ b/comfy/diffusers_convert.py
@@ -237,8 +237,12 @@ def convert_text_enc_state_dict_v20(text_enc_dict, prefix=""):
             capture_qkv_bias[k_pre][code2idx[k_code]] = v
             continue
 
-        relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k)
-        new_state_dict[relabelled_key] = v
+        text_proj = "transformer.text_projection.weight"
+        if k.endswith(text_proj):
+            new_state_dict[k.replace(text_proj, "text_projection")] = v.transpose(0, 1).contiguous()
+        else:
+            relabelled_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], k)
+            new_state_dict[relabelled_key] = v
 
     for k_pre, tensors in capture_qkv_weight.items():
         if None in tensors:
diff --git a/comfy/extra_samplers/uni_pc.py b/comfy/extra_samplers/uni_pc.py
index 08bf0fc9e67..a30d1d03f2e 100644
--- a/comfy/extra_samplers/uni_pc.py
+++ b/comfy/extra_samplers/uni_pc.py
@@ -358,9 +358,6 @@ def __init__(
         thresholding=False,
         max_val=1.,
         variant='bh1',
-        noise_mask=None,
-        masked_image=None,
-        noise=None,
     ):
         """Construct a UniPC. 
 
@@ -372,9 +369,6 @@ def __init__(
         self.predict_x0 = predict_x0
         self.thresholding = thresholding
         self.max_val = max_val
-        self.noise_mask = noise_mask
-        self.masked_image = masked_image
-        self.noise = noise
 
     def dynamic_thresholding_fn(self, x0, t=None):
         """
@@ -391,10 +385,7 @@ def noise_prediction_fn(self, x, t):
         """
         Return the noise prediction model.
         """
-        if self.noise_mask is not None:
-            return self.model(x, t) * self.noise_mask
-        else:
-            return self.model(x, t)
+        return self.model(x, t)
 
     def data_prediction_fn(self, x, t):
         """
@@ -409,8 +400,6 @@ def data_prediction_fn(self, x, t):
             s = torch.quantile(torch.abs(x0).reshape((x0.shape[0], -1)), p, dim=1)
             s = expand_dims(torch.maximum(s, self.max_val * torch.ones_like(s).to(s.device)), dims)
             x0 = torch.clamp(x0, -s, s) / s
-        if self.noise_mask is not None:
-            x0 = x0 * self.noise_mask + (1. - self.noise_mask) * self.masked_image
         return x0
 
     def model_fn(self, x, t):
@@ -723,8 +712,6 @@ def sample(self, x, timesteps, t_start=None, t_end=None, order=3, skip_type='tim
             assert timesteps.shape[0] - 1 == steps
             # with torch.no_grad():
             for step_index in trange(steps, disable=disable_pbar):
-                if self.noise_mask is not None:
-                    x = x * self.noise_mask + (1. - self.noise_mask) * (self.masked_image * self.noise_schedule.marginal_alpha(timesteps[step_index]) + self.noise * self.noise_schedule.marginal_std(timesteps[step_index]))
                 if step_index == 0:
                     vec_t = timesteps[0].expand((x.shape[0]))
                     model_prev_list = [self.model_fn(x, vec_t)]
@@ -766,7 +753,7 @@ def sample(self, x, timesteps, t_start=None, t_end=None, order=3, skip_type='tim
                                 model_x = self.model_fn(x, vec_t)
                             model_prev_list[-1] = model_x
                 if callback is not None:
-                    callback(step_index, model_prev_list[-1], x, steps)
+                    callback({'x': x, 'i': step_index, 'denoised': model_prev_list[-1]})
         else:
             raise NotImplementedError()
         # if denoise_to_zero:
@@ -858,7 +845,7 @@ def predict_eps_sigma(model, input, sigma_in, **kwargs):
     return  (input - model(input, sigma_in, **kwargs)) / sigma
 
 
-def sample_unipc(model, noise, image, sigmas, max_denoise, extra_args=None, callback=None, disable=False, noise_mask=None, variant='bh1'):
+def sample_unipc(model, noise, sigmas, extra_args=None, callback=None, disable=False, variant='bh1'):
         timesteps = sigmas.clone()
         if sigmas[-1] == 0:
             timesteps = sigmas[:]
@@ -867,16 +854,7 @@ def sample_unipc(model, noise, image, sigmas, max_denoise, extra_args=None, call
             timesteps = sigmas.clone()
         ns = SigmaConvert()
 
-        if image is not None:
-            img = image * ns.marginal_alpha(timesteps[0])
-            if max_denoise:
-                noise_mult = 1.0
-            else:
-                noise_mult = ns.marginal_std(timesteps[0])
-            img += noise * noise_mult
-        else:
-            img = noise
-
+        noise = noise / torch.sqrt(1.0 + timesteps[0] ** 2.0)
         model_type = "noise"
 
         model_fn = model_wrapper(
@@ -888,7 +866,10 @@ def sample_unipc(model, noise, image, sigmas, max_denoise, extra_args=None, call
         )
 
         order = min(3, len(timesteps) - 2)
-        uni_pc = UniPC(model_fn, ns, predict_x0=True, thresholding=False, noise_mask=noise_mask, masked_image=image, noise=noise, variant=variant)
-        x = uni_pc.sample(img, timesteps=timesteps, skip_type="time_uniform", method="multistep", order=order, lower_order_final=True, callback=callback, disable_pbar=disable)
+        uni_pc = UniPC(model_fn, ns, predict_x0=True, thresholding=False, variant=variant)
+        x = uni_pc.sample(noise, timesteps=timesteps, skip_type="time_uniform", method="multistep", order=order, lower_order_final=True, callback=callback, disable_pbar=disable)
         x /= ns.marginal_alpha(timesteps[-1])
         return x
+
+def sample_unipc_bh2(model, noise, sigmas, extra_args=None, callback=None, disable=False):
+    return sample_unipc(model, noise, sigmas, extra_args, callback, disable, variant='bh2')
\ No newline at end of file
diff --git a/comfy/latent_formats.py b/comfy/latent_formats.py
index 03fd59e3da0..674364e720f 100644
--- a/comfy/latent_formats.py
+++ b/comfy/latent_formats.py
@@ -1,3 +1,4 @@
+import torch
 
 class LatentFormat:
     scale_factor = 1.0
@@ -34,6 +35,32 @@ def __init__(self):
                 ]
         self.taesd_decoder_name = "taesdxl_decoder"
 
+class SDXL_Playground_2_5(LatentFormat):
+    def __init__(self):
+        self.scale_factor = 0.5
+        self.latents_mean = torch.tensor([-1.6574, 1.886, -1.383, 2.5155]).view(1, 4, 1, 1)
+        self.latents_std = torch.tensor([8.4927, 5.9022, 6.5498, 5.2299]).view(1, 4, 1, 1)
+
+        self.latent_rgb_factors = [
+                    #   R        G        B
+                    [ 0.3920,  0.4054,  0.4549],
+                    [-0.2634, -0.0196,  0.0653],
+                    [ 0.0568,  0.1687, -0.0755],
+                    [-0.3112, -0.2359, -0.2076]
+                ]
+        self.taesd_decoder_name = "taesdxl_decoder"
+
+    def process_in(self, latent):
+        latents_mean = self.latents_mean.to(latent.device, latent.dtype)
+        latents_std = self.latents_std.to(latent.device, latent.dtype)
+        return (latent - latents_mean) * self.scale_factor / latents_std
+
+    def process_out(self, latent):
+        latents_mean = self.latents_mean.to(latent.device, latent.dtype)
+        latents_std = self.latents_std.to(latent.device, latent.dtype)
+        return latent * latents_std / self.scale_factor + latents_mean
+
+
 class SD_X4(LatentFormat):
     def __init__(self):
         self.scale_factor = 0.08333
diff --git a/comfy/ldm/modules/diffusionmodules/openaimodel.py b/comfy/ldm/modules/diffusionmodules/openaimodel.py
index 998afd977ca..cf89ae01782 100644
--- a/comfy/ldm/modules/diffusionmodules/openaimodel.py
+++ b/comfy/ldm/modules/diffusionmodules/openaimodel.py
@@ -484,7 +484,6 @@ def __init__(
         self.predict_codebook_ids = n_embed is not None
 
         self.default_num_video_frames = None
-        self.default_image_only_indicator = None
 
         time_embed_dim = model_channels * 4
         self.time_embed = nn.Sequential(
@@ -708,27 +707,30 @@ def get_resblock(
                 device=device,
                 operations=operations
             )]
-        if transformer_depth_middle >= 0:
-            mid_block += [get_attention_layer(  # always uses a self-attn
-                            ch, num_heads, dim_head, depth=transformer_depth_middle, context_dim=context_dim,
-                            disable_self_attn=disable_middle_self_attn, use_checkpoint=use_checkpoint
-                        ),
-            get_resblock(
-                merge_factor=merge_factor,
-                merge_strategy=merge_strategy,
-                video_kernel_size=video_kernel_size,
-                ch=ch,
-                time_embed_dim=time_embed_dim,
-                dropout=dropout,
-                out_channels=None,
-                dims=dims,
-                use_checkpoint=use_checkpoint,
-                use_scale_shift_norm=use_scale_shift_norm,
-                dtype=self.dtype,
-                device=device,
-                operations=operations
-            )]
-        self.middle_block = TimestepEmbedSequential(*mid_block)
+
+        self.middle_block = None
+        if transformer_depth_middle >= -1:
+            if transformer_depth_middle >= 0:
+                mid_block += [get_attention_layer(  # always uses a self-attn
+                                ch, num_heads, dim_head, depth=transformer_depth_middle, context_dim=context_dim,
+                                disable_self_attn=disable_middle_self_attn, use_checkpoint=use_checkpoint
+                            ),
+                get_resblock(
+                    merge_factor=merge_factor,
+                    merge_strategy=merge_strategy,
+                    video_kernel_size=video_kernel_size,
+                    ch=ch,
+                    time_embed_dim=time_embed_dim,
+                    dropout=dropout,
+                    out_channels=None,
+                    dims=dims,
+                    use_checkpoint=use_checkpoint,
+                    use_scale_shift_norm=use_scale_shift_norm,
+                    dtype=self.dtype,
+                    device=device,
+                    operations=operations
+                )]
+            self.middle_block = TimestepEmbedSequential(*mid_block)
         self._feature_size += ch
 
         self.output_blocks = nn.ModuleList([])
@@ -827,7 +829,7 @@ def forward(self, x, timesteps=None, context=None, y=None, control=None, transfo
         transformer_patches = transformer_options.get("patches", {})
 
         num_video_frames = kwargs.get("num_video_frames", self.default_num_video_frames)
-        image_only_indicator = kwargs.get("image_only_indicator", self.default_image_only_indicator)
+        image_only_indicator = kwargs.get("image_only_indicator", None)
         time_context = kwargs.get("time_context", None)
 
         assert (y is not None) == (
@@ -858,7 +860,8 @@ def forward(self, x, timesteps=None, context=None, y=None, control=None, transfo
                     h = p(h, transformer_options)
 
         transformer_options["block"] = ("middle", 0)
-        h = forward_timestep_embed(self.middle_block, h, emb, context, transformer_options, time_context=time_context, num_video_frames=num_video_frames, image_only_indicator=image_only_indicator)
+        if self.middle_block is not None:
+            h = forward_timestep_embed(self.middle_block, h, emb, context, transformer_options, time_context=time_context, num_video_frames=num_video_frames, image_only_indicator=image_only_indicator)
         h = apply_control(h, control, 'middle')
 
 
diff --git a/comfy/ldm/modules/diffusionmodules/util.py b/comfy/ldm/modules/diffusionmodules/util.py
index 5a6aa7d77d1..ce14ad5e18c 100644
--- a/comfy/ldm/modules/diffusionmodules/util.py
+++ b/comfy/ldm/modules/diffusionmodules/util.py
@@ -46,23 +46,25 @@ def __init__(
         else:
             raise ValueError(f"unknown merge strategy {self.merge_strategy}")
 
-    def get_alpha(self, image_only_indicator: torch.Tensor) -> torch.Tensor:
+    def get_alpha(self, image_only_indicator: torch.Tensor, device) -> torch.Tensor:
         # skip_time_mix = rearrange(repeat(skip_time_mix, 'b -> (b t) () () ()', t=t), '(b t) 1 ... -> b 1 t ...', t=t)
         if self.merge_strategy == "fixed":
             # make shape compatible
             # alpha = repeat(self.mix_factor, '1 -> b () t  () ()', t=t, b=bs)
-            alpha = self.mix_factor.to(image_only_indicator.device)
+            alpha = self.mix_factor.to(device)
         elif self.merge_strategy == "learned":
-            alpha = torch.sigmoid(self.mix_factor.to(image_only_indicator.device))
+            alpha = torch.sigmoid(self.mix_factor.to(device))
             # make shape compatible
             # alpha = repeat(alpha, '1 -> s () ()', s = t * bs)
         elif self.merge_strategy == "learned_with_images":
-            assert image_only_indicator is not None, "need image_only_indicator ..."
-            alpha = torch.where(
-                image_only_indicator.bool(),
-                torch.ones(1, 1, device=image_only_indicator.device),
-                rearrange(torch.sigmoid(self.mix_factor.to(image_only_indicator.device)), "... -> ... 1"),
-            )
+            if image_only_indicator is None:
+                alpha = rearrange(torch.sigmoid(self.mix_factor.to(device)), "... -> ... 1")
+            else:
+                alpha = torch.where(
+                    image_only_indicator.bool(),
+                    torch.ones(1, 1, device=image_only_indicator.device),
+                    rearrange(torch.sigmoid(self.mix_factor.to(image_only_indicator.device)), "... -> ... 1"),
+                )
             alpha = rearrange(alpha, self.rearrange_pattern)
             # make shape compatible
             # alpha = repeat(alpha, '1 -> s () ()', s = t * bs)
@@ -76,7 +78,7 @@ def forward(
         x_temporal,
         image_only_indicator=None,
     ) -> torch.Tensor:
-        alpha = self.get_alpha(image_only_indicator)
+        alpha = self.get_alpha(image_only_indicator, x_spatial.device)
         x = (
             alpha.to(x_spatial.dtype) * x_spatial
             + (1.0 - alpha).to(x_spatial.dtype) * x_temporal
diff --git a/comfy/lora.py b/comfy/lora.py
index 5e4009b47f9..21b9897a63a 100644
--- a/comfy/lora.py
+++ b/comfy/lora.py
@@ -197,6 +197,15 @@ def model_lora_keys_clip(model, key_map={}):
                     key_map[lora_key] = k
                     lora_key = "text_encoder.text_model.encoder.layers.{}.{}".format(b, c) #diffusers lora
                     key_map[lora_key] = k
+                    lora_key = "lora_prior_te_text_model_encoder_layers_{}_{}".format(b, LORA_CLIP_MAP[c]) #cascade lora: TODO put lora key prefix in the model config
+                    key_map[lora_key] = k
+
+
+    k = "clip_g.transformer.text_projection.weight"
+    if k in sdk:
+        key_map["lora_prior_te_text_projection"] = k #cascade lora?
+        # key_map["text_encoder.text_projection"] = k #TODO: check if other lora have the text_projection too
+        # key_map["lora_te_text_projection"] = k
 
     return key_map
 
@@ -207,6 +216,7 @@ def model_lora_keys_unet(model, key_map={}):
         if k.startswith("diffusion_model.") and k.endswith(".weight"):
             key_lora = k[len("diffusion_model."):-len(".weight")].replace(".", "_")
             key_map["lora_unet_{}".format(key_lora)] = k
+            key_map["lora_prior_unet_{}".format(key_lora)] = k #cascade lora: TODO put lora key prefix in the model config
 
     diffusers_keys = comfy.utils.unet_to_diffusers(model.model_config.unet_config)
     for k in diffusers_keys:
diff --git a/comfy/model_base.py b/comfy/model_base.py
index 421f271b28a..a9de1366724 100644
--- a/comfy/model_base.py
+++ b/comfy/model_base.py
@@ -15,9 +15,10 @@ class ModelType(Enum):
     V_PREDICTION = 2
     V_PREDICTION_EDM = 3
     STABLE_CASCADE = 4
+    EDM = 5
 
 
-from comfy.model_sampling import EPS, V_PREDICTION, ModelSamplingDiscrete, ModelSamplingContinuousEDM, StableCascadeSampling
+from comfy.model_sampling import EPS, V_PREDICTION, EDM, ModelSamplingDiscrete, ModelSamplingContinuousEDM, StableCascadeSampling
 
 
 def model_sampling(model_config, model_type):
@@ -33,6 +34,9 @@ def model_sampling(model_config, model_type):
     elif model_type == ModelType.STABLE_CASCADE:
         c = EPS
         s = StableCascadeSampling
+    elif model_type == ModelType.EDM:
+        c = EDM
+        s = ModelSamplingContinuousEDM
 
     class ModelSampling(s, c):
         pass
@@ -163,6 +167,10 @@ def blank_inpaint_image_like(latent_image):
         if cross_attn_cnet is not None:
             out['crossattn_controlnet'] = comfy.conds.CONDCrossAttn(cross_attn_cnet)
 
+        c_concat = kwargs.get("noise_concat", None)
+        if c_concat is not None:
+            out['c_concat'] = comfy.conds.CONDNoiseShape(data)
+
         return out
 
     def load_model_weights(self, sd, unet_prefix=""):
@@ -368,7 +376,6 @@ def extra_conds(self, **kwargs):
         if "time_conditioning" in kwargs:
             out["time_context"] = comfy.conds.CONDCrossAttn(kwargs["time_conditioning"])
 
-        out['image_only_indicator'] = comfy.conds.CONDConstant(torch.zeros((1,), device=device))
         out['num_video_frames'] = comfy.conds.CONDConstant(noise.shape[0])
         return out
 
diff --git a/comfy/model_detection.py b/comfy/model_detection.py
index 8fca6d8c8e4..07ee8570864 100644
--- a/comfy/model_detection.py
+++ b/comfy/model_detection.py
@@ -151,8 +151,10 @@ def detect_unet_config(state_dict, key_prefix):
     channel_mult.append(last_channel_mult)
     if "{}middle_block.1.proj_in.weight".format(key_prefix) in state_dict_keys:
         transformer_depth_middle = count_blocks(state_dict_keys, '{}middle_block.1.transformer_blocks.'.format(key_prefix) + '{}')
-    else:
+    elif "{}middle_block.0.in_layers.0.weight".format(key_prefix) in state_dict_keys:
         transformer_depth_middle = -1
+    else:
+        transformer_depth_middle = -2
 
     unet_config["in_channels"] = in_channels
     unet_config["out_channels"] = out_channels
@@ -242,6 +244,7 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
     down_blocks = count_blocks(state_dict, "down_blocks.{}")
     for i in range(down_blocks):
         attn_blocks = count_blocks(state_dict, "down_blocks.{}.attentions.".format(i) + '{}')
+        res_blocks = count_blocks(state_dict, "down_blocks.{}.resnets.".format(i) + '{}')
         for ab in range(attn_blocks):
             transformer_count = count_blocks(state_dict, "down_blocks.{}.attentions.{}.transformer_blocks.".format(i, ab) + '{}')
             transformer_depth.append(transformer_count)
@@ -250,8 +253,8 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
 
         attn_res *= 2
         if attn_blocks == 0:
-            transformer_depth.append(0)
-            transformer_depth.append(0)
+            for i in range(res_blocks):
+                transformer_depth.append(0)
 
     match["transformer_depth"] = transformer_depth
 
@@ -329,7 +332,19 @@ def unet_config_from_diffusers_unet(state_dict, dtype=None):
               'channel_mult': [1, 2, 4], 'transformer_depth_middle': -1, 'use_linear_in_transformer': True, 'context_dim': 2048, 'num_head_channels': 64,
               'use_temporal_attention': False, 'use_temporal_resblock': False}
 
-    supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega]
+    KOALA_700M = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False,
+              'num_classes': 'sequential', 'adm_in_channels': 2816, 'dtype': dtype, 'in_channels': 4, 'model_channels': 320,
+              'num_res_blocks': [1, 1, 1], 'transformer_depth': [0, 2, 5], 'transformer_depth_output': [0, 0, 2, 2, 5, 5],
+              'channel_mult': [1, 2, 4], 'transformer_depth_middle': -2, 'use_linear_in_transformer': True, 'context_dim': 2048, 'num_head_channels': 64,
+              'use_temporal_attention': False, 'use_temporal_resblock': False}
+
+    KOALA_1B = {'use_checkpoint': False, 'image_size': 32, 'out_channels': 4, 'use_spatial_transformer': True, 'legacy': False,
+              'num_classes': 'sequential', 'adm_in_channels': 2816, 'dtype': dtype, 'in_channels': 4, 'model_channels': 320,
+              'num_res_blocks': [1, 1, 1], 'transformer_depth': [0, 2, 6], 'transformer_depth_output': [0, 0, 2, 2, 6, 6],
+              'channel_mult': [1, 2, 4], 'transformer_depth_middle': 6, 'use_linear_in_transformer': True, 'context_dim': 2048, 'num_head_channels': 64,
+              'use_temporal_attention': False, 'use_temporal_resblock': False}
+
+    supported_models = [SDXL, SDXL_refiner, SD21, SD15, SD21_uncliph, SD21_unclipl, SDXL_mid_cnet, SDXL_small_cnet, SDXL_diffusers_inpaint, SSD_1B, Segmind_Vega, KOALA_700M, KOALA_1B]
 
     for unet_config in supported_models:
         matches = True
diff --git a/comfy/model_management.py b/comfy/model_management.py
index d2d8a4d06d8..f0d726fcdae 100644
--- a/comfy/model_management.py
+++ b/comfy/model_management.py
@@ -767,7 +767,7 @@ def should_use_fp16(device=None, model_params=0, prioritize_performance=True, ma
     #FP16 is confirmed working on a 1080 (GP104) but it's a bit slower than FP32 so it should only be enabled
     #when the model doesn't actually fit on the card
     #TODO: actually test if GP106 and others have the same type of behavior
-    nvidia_10_series = ["1080", "1070", "titan x", "p3000", "p3200", "p4000", "p4200", "p5000", "p5200", "p6000", "1060", "1050"]
+    nvidia_10_series = ["1080", "1070", "titan x", "p3000", "p3200", "p4000", "p4200", "p5000", "p5200", "p6000", "1060", "1050", "p40", "p100", "p6", "p4"]
     for x in nvidia_10_series:
         if x in props.name.lower():
             fp16_works = True
diff --git a/comfy/model_patcher.py b/comfy/model_patcher.py
index a88b737cca3..4a5d42b035c 100644
--- a/comfy/model_patcher.py
+++ b/comfy/model_patcher.py
@@ -67,6 +67,9 @@ def set_model_sampler_post_cfg_function(self, post_cfg_function, disable_cfg1_op
     def set_model_unet_function_wrapper(self, unet_wrapper_function):
         self.model_options["model_function_wrapper"] = unet_wrapper_function
 
+    def set_model_denoise_mask_function(self, denoise_mask_function):
+        self.model_options["denoise_mask_function"] = denoise_mask_function
+
     def set_model_patch(self, patch, name):
         to = self.model_options["transformer_options"]
         if "patches" not in to:
@@ -176,10 +179,9 @@ def model_state_dict(self, filter_prefix=None):
 
     def patch_model(self, device_to=None, patch_weights=True):
         for k in self.object_patches:
-            old = getattr(self.model, k)
+            old = comfy.utils.set_attr(self.model, k, self.object_patches[k])
             if k not in self.object_patches_backup:
                 self.object_patches_backup[k] = old
-            setattr(self.model, k, self.object_patches[k])
 
         if patch_weights:
             model_sd = self.model_state_dict()
@@ -203,7 +205,7 @@ def patch_model(self, device_to=None, patch_weights=True):
                 if inplace_update:
                     comfy.utils.copy_to_param(self.model, key, out_weight)
                 else:
-                    comfy.utils.set_attr(self.model, key, out_weight)
+                    comfy.utils.set_attr_param(self.model, key, out_weight)
                 del temp_weight
 
             if device_to is not None:
@@ -342,7 +344,7 @@ def unpatch_model(self, device_to=None):
                 comfy.utils.copy_to_param(self.model, k, self.backup[k])
         else:
             for k in keys:
-                comfy.utils.set_attr(self.model, k, self.backup[k])
+                comfy.utils.set_attr_param(self.model, k, self.backup[k])
 
         self.backup = {}
 
@@ -352,6 +354,6 @@ def unpatch_model(self, device_to=None):
 
         keys = list(self.object_patches_backup.keys())
         for k in keys:
-            setattr(self.model, k, self.object_patches_backup[k])
+            comfy.utils.set_attr(self.model, k, self.object_patches_backup[k])
 
         self.object_patches_backup = {}
diff --git a/comfy/model_sampling.py b/comfy/model_sampling.py
index 97e91a01d67..b46202b7803 100644
--- a/comfy/model_sampling.py
+++ b/comfy/model_sampling.py
@@ -11,12 +11,25 @@ def calculate_denoised(self, sigma, model_output, model_input):
         sigma = sigma.view(sigma.shape[:1] + (1,) * (model_output.ndim - 1))
         return model_input - model_output * sigma
 
+    def noise_scaling(self, sigma, noise, latent_image, max_denoise=False):
+        if max_denoise:
+            noise = noise * torch.sqrt(1.0 + sigma ** 2.0)
+        else:
+            noise = noise * sigma
+        if latent_image is not None:
+            noise += latent_image
+        return noise
 
 class V_PREDICTION(EPS):
     def calculate_denoised(self, sigma, model_output, model_input):
         sigma = sigma.view(sigma.shape[:1] + (1,) * (model_output.ndim - 1))
         return model_input * self.sigma_data ** 2 / (sigma ** 2 + self.sigma_data ** 2) - model_output * sigma * self.sigma_data / (sigma ** 2 + self.sigma_data ** 2) ** 0.5
 
+class EDM(V_PREDICTION):
+    def calculate_denoised(self, sigma, model_output, model_input):
+        sigma = sigma.view(sigma.shape[:1] + (1,) * (model_output.ndim - 1))
+        return model_input * self.sigma_data ** 2 / (sigma ** 2 + self.sigma_data ** 2) + model_output * sigma * self.sigma_data / (sigma ** 2 + self.sigma_data ** 2) ** 0.5
+
 
 class ModelSamplingDiscrete(torch.nn.Module):
     def __init__(self, model_config=None):
@@ -92,8 +105,6 @@ def percent_to_sigma(self, percent):
 class ModelSamplingContinuousEDM(torch.nn.Module):
     def __init__(self, model_config=None):
         super().__init__()
-        self.sigma_data = 1.0
-
         if model_config is not None:
             sampling_settings = model_config.sampling_settings
         else:
@@ -101,9 +112,11 @@ def __init__(self, model_config=None):
 
         sigma_min = sampling_settings.get("sigma_min", 0.002)
         sigma_max = sampling_settings.get("sigma_max", 120.0)
-        self.set_sigma_range(sigma_min, sigma_max)
+        sigma_data = sampling_settings.get("sigma_data", 1.0)
+        self.set_parameters(sigma_min, sigma_max, sigma_data)
 
-    def set_sigma_range(self, sigma_min, sigma_max):
+    def set_parameters(self, sigma_min, sigma_max, sigma_data):
+        self.sigma_data = sigma_data
         sigmas = torch.linspace(math.log(sigma_min), math.log(sigma_max), 1000).exp()
 
         self.register_buffer('sigmas', sigmas) #for compatibility with some schedulers
diff --git a/comfy/samplers.py b/comfy/samplers.py
index c795f208d80..6863be4eb7b 100644
--- a/comfy/samplers.py
+++ b/comfy/samplers.py
@@ -208,6 +208,7 @@ def calc_cond_uncond_batch(model, cond, uncond, x_in, timestep, model_options):
                         cur_patches[p] = cur_patches[p] + patches[p]
                     else:
                         cur_patches[p] = patches[p]
+                transformer_options["patches"] = cur_patches
             else:
                 transformer_options["patches"] = patches
 
@@ -271,13 +272,16 @@ def forward(self, *args, **kwargs):
         return self.apply_model(*args, **kwargs)
 
 class KSamplerX0Inpaint(torch.nn.Module):
-    def __init__(self, model):
+    def __init__(self, model, sigmas):
         super().__init__()
         self.inner_model = model
+        self.sigmas = sigmas
     def forward(self, x, sigma, uncond, cond, cond_scale, denoise_mask, model_options={}, seed=None):
         if denoise_mask is not None:
+            if "denoise_mask_function" in model_options:
+                denoise_mask = model_options["denoise_mask_function"](sigma, denoise_mask, extra_options={"model": self.inner_model, "sigmas": self.sigmas})
             latent_mask = 1. - denoise_mask
-            x = x * denoise_mask + (self.latent_image + self.noise * sigma.reshape([sigma.shape[0]] + [1] * (len(self.noise.shape) - 1))) * latent_mask
+            x = x * denoise_mask + self.inner_model.inner_model.model_sampling.noise_scaling(sigma.reshape([sigma.shape[0]] + [1] * (len(self.noise.shape) - 1)), self.noise, self.latent_image) * latent_mask
         out = self.inner_model(x, sigma, cond=cond, uncond=uncond, cond_scale=cond_scale, model_options=model_options, seed=seed)
         if denoise_mask is not None:
             out = out * denoise_mask + self.latent_image * latent_mask
@@ -513,14 +517,6 @@ def max_denoise(self, model_wrap, sigmas):
         sigma = float(sigmas[0])
         return math.isclose(max_sigma, sigma, rel_tol=1e-05) or sigma > max_sigma
 
-class UNIPC(Sampler):
-    def sample(self, model_wrap, sigmas, extra_args, callback, noise, latent_image=None, denoise_mask=None, disable_pbar=False):
-        return uni_pc.sample_unipc(model_wrap, noise, latent_image, sigmas, max_denoise=self.max_denoise(model_wrap, sigmas), extra_args=extra_args, noise_mask=denoise_mask, callback=callback, disable=disable_pbar)
-
-class UNIPCBH2(Sampler):
-    def sample(self, model_wrap, sigmas, extra_args, callback, noise, latent_image=None, denoise_mask=None, disable_pbar=False):
-        return uni_pc.sample_unipc(model_wrap, noise, latent_image, sigmas, max_denoise=self.max_denoise(model_wrap, sigmas), extra_args=extra_args, noise_mask=denoise_mask, callback=callback, variant='bh2', disable=disable_pbar)
-
 KSAMPLER_NAMES = ["euler", "euler_ancestral", "heun", "heunpp2","dpm_2", "dpm_2_ancestral",
                   "lms", "dpm_fast", "dpm_adaptive", "dpmpp_2s_ancestral", "dpmpp_sde", "dpmpp_sde_gpu",
                   "dpmpp_2m", "dpmpp_2m_sde", "dpmpp_2m_sde_gpu", "dpmpp_3m_sde", "dpmpp_3m_sde_gpu", "ddpm", "lcm"]
@@ -533,7 +529,7 @@ def __init__(self, sampler_function, extra_options={}, inpaint_options={}):
 
     def sample(self, model_wrap, sigmas, extra_args, callback, noise, latent_image=None, denoise_mask=None, disable_pbar=False):
         extra_args["denoise_mask"] = denoise_mask
-        model_k = KSamplerX0Inpaint(model_wrap)
+        model_k = KSamplerX0Inpaint(model_wrap, sigmas)
         model_k.latent_image = latent_image
         if self.inpaint_options.get("random", False): #TODO: Should this be the default?
             generator = torch.manual_seed(extra_args.get("seed", 41) + 1)
@@ -541,19 +537,13 @@ def sample(self, model_wrap, sigmas, extra_args, callback, noise, latent_image=N
         else:
             model_k.noise = noise
 
-        if self.max_denoise(model_wrap, sigmas):
-            noise = noise * torch.sqrt(1.0 + sigmas[0] ** 2.0)
-        else:
-            noise = noise * sigmas[0]
+        noise = model_wrap.inner_model.model_sampling.noise_scaling(sigmas[0], noise, latent_image, self.max_denoise(model_wrap, sigmas))
 
         k_callback = None
         total_steps = len(sigmas) - 1
         if callback is not None:
             k_callback = lambda x: callback(x["i"], x["denoised"], x["x"], total_steps)
 
-        if latent_image is not None:
-            noise += latent_image
-
         samples = self.sampler_function(model_k, noise, sigmas, extra_args=extra_args, callback=k_callback, disable=disable_pbar, **self.extra_options)
         return samples
 
@@ -595,7 +585,7 @@ def sample(model, noise, positive, negative, cfg, device, sampler, sigmas, model
     calculate_start_end_timesteps(model, negative)
     calculate_start_end_timesteps(model, positive)
 
-    if latent_image is not None:
+    if latent_image is not None and torch.count_nonzero(latent_image) > 0: #Don't shift the empty latent image.
         latent_image = model.process_latent_in(latent_image)
 
     if hasattr(model, 'extra_conds'):
@@ -640,9 +630,9 @@ def calculate_sigmas_scheduler(model, scheduler_name, steps):
 
 def sampler_object(name):
     if name == "uni_pc":
-        sampler = UNIPC()
+        sampler = KSAMPLER(uni_pc.sample_unipc)
     elif name == "uni_pc_bh2":
-        sampler = UNIPCBH2()
+        sampler = KSAMPLER(uni_pc.sample_unipc_bh2)
     elif name == "ddim":
         sampler = ksampler("euler", inpaint_options={"random": True})
     else:
diff --git a/comfy/sd.py b/comfy/sd.py
index 48d567009a3..bcc900558f9 100644
--- a/comfy/sd.py
+++ b/comfy/sd.py
@@ -52,7 +52,7 @@ def load_clip_weights(model, sd):
         if ids.dtype == torch.float32:
             sd['cond_stage_model.transformer.text_model.embeddings.position_ids'] = ids.round()
 
-    sd = comfy.utils.transformers_convert(sd, "cond_stage_model.model.", "cond_stage_model.transformer.text_model.", 24)
+    sd = comfy.utils.clip_text_transformers_convert(sd, "cond_stage_model.model.", "cond_stage_model.transformer.")
     return load_model_weights(model, sd)
 
 
@@ -123,10 +123,13 @@ def tokenize(self, text, return_word_ids=False):
         return self.tokenizer.tokenize_with_weights(text, return_word_ids)
 
     def encode_from_tokens(self, tokens, return_pooled=False):
+        self.cond_stage_model.reset_clip_options()
+
         if self.layer_idx is not None:
-            self.cond_stage_model.clip_layer(self.layer_idx)
-        else:
-            self.cond_stage_model.reset_clip_layer()
+            self.cond_stage_model.set_clip_options({"layer": self.layer_idx})
+
+        if return_pooled == "unprojected":
+            self.cond_stage_model.set_clip_options({"projected_pooled": False})
 
         self.load_model()
         cond, pooled = self.cond_stage_model.encode_token_weights(tokens)
@@ -391,7 +394,10 @@ class EmptyClass:
 
     for i in range(len(clip_data)):
         if "transformer.resblocks.0.ln_1.weight" in clip_data[i]:
-            clip_data[i] = comfy.utils.transformers_convert(clip_data[i], "", "text_model.", 32)
+            clip_data[i] = comfy.utils.clip_text_transformers_convert(clip_data[i], "", "")
+        else:
+            if "text_projection" in clip_data[i]:
+                clip_data[i]["text_projection.weight"] = clip_data[i]["text_projection"].transpose(0, 1) #old models saved with the CLIPSave node
 
     clip_target = EmptyClass()
     clip_target.params = {}
diff --git a/comfy/sd1_clip.py b/comfy/sd1_clip.py
index 8287ad2e8b8..87e3eaa4ddb 100644
--- a/comfy/sd1_clip.py
+++ b/comfy/sd1_clip.py
@@ -67,7 +67,7 @@ class SDClipModel(torch.nn.Module, ClipTokenWeightEncoder):
     ]
     def __init__(self, version="openai/clip-vit-large-patch14", device="cpu", max_length=77,
                  freeze=True, layer="last", layer_idx=None, textmodel_json_config=None, dtype=None, model_class=comfy.clip_model.CLIPTextModel,
-                 special_tokens={"start": 49406, "end": 49407, "pad": 49407}, layer_norm_hidden_state=True, enable_attention_masks=False):  # clip-vit-base-patch32
+                 special_tokens={"start": 49406, "end": 49407, "pad": 49407}, layer_norm_hidden_state=True, enable_attention_masks=False, return_projected_pooled=True):  # clip-vit-base-patch32
         super().__init__()
         assert layer in self.LAYERS
 
@@ -86,16 +86,18 @@ def __init__(self, version="openai/clip-vit-large-patch14", device="cpu", max_le
         self.layer = layer
         self.layer_idx = None
         self.special_tokens = special_tokens
-        self.text_projection = torch.nn.Parameter(torch.eye(self.transformer.get_input_embeddings().weight.shape[1]))
+
         self.logit_scale = torch.nn.Parameter(torch.tensor(4.6055))
         self.enable_attention_masks = enable_attention_masks
 
         self.layer_norm_hidden_state = layer_norm_hidden_state
+        self.return_projected_pooled = return_projected_pooled
+
         if layer == "hidden":
             assert layer_idx is not None
             assert abs(layer_idx) < self.num_layers
-            self.clip_layer(layer_idx)
-        self.layer_default = (self.layer, self.layer_idx)
+            self.set_clip_options({"layer": layer_idx})
+        self.options_default = (self.layer, self.layer_idx, self.return_projected_pooled)
 
     def freeze(self):
         self.transformer = self.transformer.eval()
@@ -103,16 +105,19 @@ def freeze(self):
         for param in self.parameters():
             param.requires_grad = False
 
-    def clip_layer(self, layer_idx):
-        if abs(layer_idx) > self.num_layers:
+    def set_clip_options(self, options):
+        layer_idx = options.get("layer", self.layer_idx)
+        self.return_projected_pooled = options.get("projected_pooled", self.return_projected_pooled)
+        if layer_idx is None or abs(layer_idx) > self.num_layers:
             self.layer = "last"
         else:
             self.layer = "hidden"
             self.layer_idx = layer_idx
 
-    def reset_clip_layer(self):
-        self.layer = self.layer_default[0]
-        self.layer_idx = self.layer_default[1]
+    def reset_clip_options(self):
+        self.layer = self.options_default[0]
+        self.layer_idx = self.options_default[1]
+        self.return_projected_pooled = self.options_default[2]
 
     def set_up_textual_embeddings(self, tokens, current_embeds):
         out_tokens = []
@@ -177,23 +182,19 @@ def forward(self, tokens):
         else:
             z = outputs[1]
 
-        if outputs[2] is not None:
-            pooled_output = outputs[2].float()
-        else:
-            pooled_output = None
+        pooled_output = None
+        if len(outputs) >= 3:
+            if not self.return_projected_pooled and len(outputs) >= 4 and outputs[3] is not None:
+                pooled_output = outputs[3].float()
+            elif outputs[2] is not None:
+                pooled_output = outputs[2].float()
 
-        if self.text_projection is not None and pooled_output is not None:
-            pooled_output = pooled_output.float().to(self.text_projection.device) @ self.text_projection.float()
         return z.float(), pooled_output
 
     def encode(self, tokens):
         return self(tokens)
 
     def load_sd(self, sd):
-        if "text_projection" in sd:
-            self.text_projection[:] = sd.pop("text_projection")
-        if "text_projection.weight" in sd:
-            self.text_projection[:] = sd.pop("text_projection.weight").transpose(0, 1)
         return self.transformer.load_state_dict(sd, strict=False)
 
 def parse_parentheses(string):
@@ -354,11 +355,12 @@ def load_embed(embedding_name, embedding_directory, embedding_size, embed_key=No
     return embed_out
 
 class SDTokenizer:
-    def __init__(self, tokenizer_path=None, max_length=77, pad_with_end=True, embedding_directory=None, embedding_size=768, embedding_key='clip_l', tokenizer_class=CLIPTokenizer, has_start_token=True, pad_to_max_length=True):
+    def __init__(self, tokenizer_path=None, max_length=77, pad_with_end=True, embedding_directory=None, embedding_size=768, embedding_key='clip_l', tokenizer_class=CLIPTokenizer, has_start_token=True, pad_to_max_length=True, min_length=None):
         if tokenizer_path is None:
             tokenizer_path = os.path.join(os.path.dirname(os.path.realpath(__file__)), "sd1_tokenizer")
         self.tokenizer = tokenizer_class.from_pretrained(tokenizer_path)
         self.max_length = max_length
+        self.min_length = min_length
 
         empty = self.tokenizer('')["input_ids"]
         if has_start_token:
@@ -470,6 +472,8 @@ def tokenize_with_weights(self, text:str, return_word_ids=False):
         batch.append((self.end_token, 1.0, 0))
         if self.pad_to_max_length:
             batch.extend([(pad_token, 1.0, 0)] * (self.max_length - len(batch)))
+        if self.min_length is not None and len(batch) < self.min_length:
+            batch.extend([(pad_token, 1.0, 0)] * (self.min_length - len(batch)))
 
         if not return_word_ids:
             batched_tokens = [[(t, w) for t, w,_ in x] for x in batched_tokens]
@@ -503,11 +507,11 @@ def __init__(self, device="cpu", dtype=None, clip_name="l", clip_model=SDClipMod
         self.clip = "clip_{}".format(self.clip_name)
         setattr(self, self.clip, clip_model(device=device, dtype=dtype, **kwargs))
 
-    def clip_layer(self, layer_idx):
-        getattr(self, self.clip).clip_layer(layer_idx)
+    def set_clip_options(self, options):
+        getattr(self, self.clip).set_clip_options(options)
 
-    def reset_clip_layer(self):
-        getattr(self, self.clip).reset_clip_layer()
+    def reset_clip_options(self):
+        getattr(self, self.clip).reset_clip_options()
 
     def encode_token_weights(self, token_weight_pairs):
         token_weight_pairs = token_weight_pairs[self.clip_name]
diff --git a/comfy/sdxl_clip.py b/comfy/sdxl_clip.py
index 3ce5c7e05e6..e62d1ed868c 100644
--- a/comfy/sdxl_clip.py
+++ b/comfy/sdxl_clip.py
@@ -40,13 +40,13 @@ def __init__(self, device="cpu", dtype=None):
         self.clip_l = sd1_clip.SDClipModel(layer="hidden", layer_idx=-2, device=device, dtype=dtype, layer_norm_hidden_state=False)
         self.clip_g = SDXLClipG(device=device, dtype=dtype)
 
-    def clip_layer(self, layer_idx):
-        self.clip_l.clip_layer(layer_idx)
-        self.clip_g.clip_layer(layer_idx)
+    def set_clip_options(self, options):
+        self.clip_l.set_clip_options(options)
+        self.clip_g.set_clip_options(options)
 
-    def reset_clip_layer(self):
-        self.clip_g.reset_clip_layer()
-        self.clip_l.reset_clip_layer()
+    def reset_clip_options(self):
+        self.clip_g.reset_clip_options()
+        self.clip_l.reset_clip_options()
 
     def encode_token_weights(self, token_weight_pairs):
         token_weight_pairs_g = token_weight_pairs["g"]
diff --git a/comfy/supported_models.py b/comfy/supported_models.py
index 5bb98d88a96..3758210326c 100644
--- a/comfy/supported_models.py
+++ b/comfy/supported_models.py
@@ -75,7 +75,7 @@ def process_clip_state_dict(self, state_dict):
         replace_prefix["conditioner.embedders.0.model."] = "clip_h." #SD2 in sgm format
         replace_prefix["cond_stage_model.model."] = "clip_h."
         state_dict = utils.state_dict_prefix_replace(state_dict, replace_prefix, filter_keys=True)
-        state_dict = utils.transformers_convert(state_dict, "clip_h.", "clip_h.transformer.text_model.", 24)
+        state_dict = utils.clip_text_transformers_convert(state_dict, "clip_h.", "clip_h.transformer.")
         return state_dict
 
     def process_clip_state_dict_for_saving(self, state_dict):
@@ -134,7 +134,7 @@ def process_clip_state_dict(self, state_dict):
         replace_prefix["conditioner.embedders.0.model."] = "clip_g."
         state_dict = utils.state_dict_prefix_replace(state_dict, replace_prefix, filter_keys=True)
 
-        state_dict = utils.transformers_convert(state_dict, "clip_g.", "clip_g.transformer.text_model.", 32)
+        state_dict = utils.clip_text_transformers_convert(state_dict, "clip_g.", "clip_g.transformer.")
         state_dict = utils.state_dict_key_replace(state_dict, keys_to_replace)
         return state_dict
 
@@ -163,7 +163,13 @@ class SDXL(supported_models_base.BASE):
     latent_format = latent_formats.SDXL
 
     def model_type(self, state_dict, prefix=""):
-        if "v_pred" in state_dict:
+        if 'edm_mean' in state_dict and 'edm_std' in state_dict: #Playground V2.5
+            self.latent_format = latent_formats.SDXL_Playground_2_5()
+            self.sampling_settings["sigma_data"] = 0.5
+            self.sampling_settings["sigma_max"] = 80.0
+            self.sampling_settings["sigma_min"] = 0.002
+            return model_base.ModelType.EDM
+        elif "v_pred" in state_dict:
             return model_base.ModelType.V_PREDICTION
         else:
             return model_base.ModelType.EPS
@@ -182,22 +188,24 @@ def process_clip_state_dict(self, state_dict):
         replace_prefix["conditioner.embedders.1.model."] = "clip_g."
         state_dict = utils.state_dict_prefix_replace(state_dict, replace_prefix, filter_keys=True)
 
-        state_dict = utils.transformers_convert(state_dict, "clip_g.", "clip_g.transformer.text_model.", 32)
-        keys_to_replace["clip_g.text_projection.weight"] = "clip_g.text_projection"
-
         state_dict = utils.state_dict_key_replace(state_dict, keys_to_replace)
+        state_dict = utils.clip_text_transformers_convert(state_dict, "clip_g.", "clip_g.transformer.")
         return state_dict
 
     def process_clip_state_dict_for_saving(self, state_dict):
         replace_prefix = {}
         keys_to_replace = {}
         state_dict_g = diffusers_convert.convert_text_enc_state_dict_v20(state_dict, "clip_g")
-        if "clip_g.transformer.text_model.embeddings.position_ids" in state_dict_g:
-            state_dict_g.pop("clip_g.transformer.text_model.embeddings.position_ids")
         for k in state_dict:
             if k.startswith("clip_l"):
                 state_dict_g[k] = state_dict[k]
 
+        state_dict_g["clip_l.transformer.text_model.embeddings.position_ids"] = torch.arange(77).expand((1, -1))
+        pop_keys = ["clip_l.transformer.text_projection.weight", "clip_l.logit_scale"]
+        for p in pop_keys:
+            if p in state_dict_g:
+                state_dict_g.pop(p)
+
         replace_prefix["clip_g"] = "conditioner.embedders.1.model"
         replace_prefix["clip_l"] = "conditioner.embedders.0"
         state_dict_g = utils.state_dict_prefix_replace(state_dict_g, replace_prefix)
@@ -226,6 +234,26 @@ class Segmind_Vega(SDXL):
         "use_temporal_attention": False,
     }
 
+class KOALA_700M(SDXL):
+    unet_config = {
+        "model_channels": 320,
+        "use_linear_in_transformer": True,
+        "transformer_depth": [0, 2, 5],
+        "context_dim": 2048,
+        "adm_in_channels": 2816,
+        "use_temporal_attention": False,
+    }
+
+class KOALA_1B(SDXL):
+    unet_config = {
+        "model_channels": 320,
+        "use_linear_in_transformer": True,
+        "transformer_depth": [0, 2, 6],
+        "context_dim": 2048,
+        "adm_in_channels": 2816,
+        "use_temporal_attention": False,
+    }
+
 class SVD_img2vid(supported_models_base.BASE):
     unet_config = {
         "model_channels": 320,
@@ -338,6 +366,12 @@ def process_unet_state_dict(self, state_dict):
                     state_dict[k_to] = weights[shape_from*x:shape_from*(x + 1)]
         return state_dict
 
+    def process_clip_state_dict(self, state_dict):
+        state_dict = utils.state_dict_prefix_replace(state_dict, {k: "" for k in self.text_encoder_key_prefix}, filter_keys=True)
+        if "clip_g.text_projection" in state_dict:
+            state_dict["clip_g.transformer.text_projection.weight"] = state_dict.pop("clip_g.text_projection").transpose(0, 1)
+        return state_dict
+
     def get_model(self, state_dict, prefix="", device=None):
         out = model_base.StableCascade_C(self, device=device)
         return out
@@ -366,5 +400,5 @@ def get_model(self, state_dict, prefix="", device=None):
         return out
 
 
-models = [Stable_Zero123, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXLRefiner, SDXL, SSD1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B]
+models = [Stable_Zero123, SD15, SD20, SD21UnclipL, SD21UnclipH, SDXLRefiner, SDXL, SSD1B, KOALA_700M, KOALA_1B, Segmind_Vega, SD_X4Upscaler, Stable_Cascade_C, Stable_Cascade_B]
 models += [SVD_img2vid]
diff --git a/comfy/utils.py b/comfy/utils.py
index 1ccdda638a0..819b92badb5 100644
--- a/comfy/utils.py
+++ b/comfy/utils.py
@@ -124,8 +124,22 @@ def transformers_convert(sd, prefix_from, prefix_to, number):
                     p = ["self_attn.q_proj", "self_attn.k_proj", "self_attn.v_proj"]
                     k_to = "{}encoder.layers.{}.{}.{}".format(prefix_to, resblock, p[x], y)
                     sd[k_to] = weights[shape_from*x:shape_from*(x + 1)]
+
     return sd
 
+def clip_text_transformers_convert(sd, prefix_from, prefix_to):
+    sd = transformers_convert(sd, prefix_from, "{}text_model.".format(prefix_to), 32)
+
+    tp = "{}text_projection.weight".format(prefix_from)
+    if tp in sd:
+        sd["{}text_projection.weight".format(prefix_to)] = sd.pop(tp)
+
+    tp = "{}text_projection".format(prefix_from)
+    if tp in sd:
+        sd["{}text_projection.weight".format(prefix_to)] = sd.pop(tp).transpose(0, 1).contiguous()
+    return sd
+
+
 UNET_MAP_ATTENTIONS = {
     "proj_in.weight",
     "proj_in.bias",
@@ -306,8 +320,11 @@ def set_attr(obj, attr, value):
     for name in attrs[:-1]:
         obj = getattr(obj, name)
     prev = getattr(obj, attrs[-1])
-    setattr(obj, attrs[-1], torch.nn.Parameter(value, requires_grad=False))
-    del prev
+    setattr(obj, attrs[-1], value)
+    return prev
+
+def set_attr_param(obj, attr, value):
+    return set_attr(obj, attr, torch.nn.Parameter(value, requires_grad=False))
 
 def copy_to_param(obj, attr, value):
     # inplace update tensor instead of replacing it
diff --git a/comfy_extras/nodes_differential_diffusion.py b/comfy_extras/nodes_differential_diffusion.py
new file mode 100644
index 00000000000..7e858a71b08
--- /dev/null
+++ b/comfy_extras/nodes_differential_diffusion.py
@@ -0,0 +1,42 @@
+# code adapted from https://github.com/exx8/differential-diffusion
+
+import torch
+
+class DifferentialDiffusion():
+    @classmethod
+    def INPUT_TYPES(s):
+        return {"required": {"model": ("MODEL", ),
+                            }}
+    RETURN_TYPES = ("MODEL",)
+    FUNCTION = "apply"
+    CATEGORY = "_for_testing"
+    INIT = False
+
+    def apply(self, model):
+        model = model.clone()
+        model.set_model_denoise_mask_function(self.forward)
+        return (model,)
+
+    def forward(self, sigma: torch.Tensor, denoise_mask: torch.Tensor, extra_options: dict):
+        model = extra_options["model"]
+        step_sigmas = extra_options["sigmas"]
+        sigma_to = model.inner_model.model_sampling.sigma_min
+        if step_sigmas[-1] > sigma_to:
+            sigma_to = step_sigmas[-1]
+        sigma_from = step_sigmas[0]
+
+        ts_from = model.inner_model.model_sampling.timestep(sigma_from)
+        ts_to = model.inner_model.model_sampling.timestep(sigma_to)
+        current_ts = model.inner_model.model_sampling.timestep(sigma)
+
+        threshold = (current_ts - ts_to) / (ts_from - ts_to)
+
+        return (denoise_mask >= threshold).to(denoise_mask.dtype)
+
+
+NODE_CLASS_MAPPINGS = {
+    "DifferentialDiffusion": DifferentialDiffusion,
+}
+NODE_DISPLAY_NAME_MAPPINGS = {
+    "DifferentialDiffusion": "Differential Diffusion",
+}
diff --git a/comfy_extras/nodes_model_advanced.py b/comfy_extras/nodes_model_advanced.py
index 1b3f3945e38..21af4b73339 100644
--- a/comfy_extras/nodes_model_advanced.py
+++ b/comfy_extras/nodes_model_advanced.py
@@ -1,6 +1,7 @@
 import folder_paths
 import comfy.sd
 import comfy.model_sampling
+import comfy.latent_formats
 import torch
 
 class LCM(comfy.model_sampling.EPS):
@@ -135,7 +136,7 @@ class ModelSamplingContinuousEDM:
     @classmethod
     def INPUT_TYPES(s):
         return {"required": { "model": ("MODEL",),
-                              "sampling": (["v_prediction", "eps"],),
+                              "sampling": (["v_prediction", "edm_playground_v2.5", "eps"],),
                               "sigma_max": ("FLOAT", {"default": 120.0, "min": 0.0, "max": 1000.0, "step":0.001, "round": False}),
                               "sigma_min": ("FLOAT", {"default": 0.002, "min": 0.0, "max": 1000.0, "step":0.001, "round": False}),
                               }}
@@ -148,17 +149,25 @@ def INPUT_TYPES(s):
     def patch(self, model, sampling, sigma_max, sigma_min):
         m = model.clone()
 
+        latent_format = None
+        sigma_data = 1.0
         if sampling == "eps":
             sampling_type = comfy.model_sampling.EPS
         elif sampling == "v_prediction":
             sampling_type = comfy.model_sampling.V_PREDICTION
+        elif sampling == "edm_playground_v2.5":
+            sampling_type = comfy.model_sampling.EDM
+            sigma_data = 0.5
+            latent_format = comfy.latent_formats.SDXL_Playground_2_5()
 
         class ModelSamplingAdvanced(comfy.model_sampling.ModelSamplingContinuousEDM, sampling_type):
             pass
 
         model_sampling = ModelSamplingAdvanced(model.model.model_config)
-        model_sampling.set_sigma_range(sigma_min, sigma_max)
+        model_sampling.set_parameters(sigma_min, sigma_max, sigma_data)
         m.add_object_patch("model_sampling", model_sampling)
+        if latent_format is not None:
+            m.add_object_patch("latent_format", latent_format)
         return (m, )
 
 class RescaleCFG:
diff --git a/comfy_extras/nodes_perpneg.py b/comfy_extras/nodes_perpneg.py
index 45e4d418f4f..64bbc1dcd42 100644
--- a/comfy_extras/nodes_perpneg.py
+++ b/comfy_extras/nodes_perpneg.py
@@ -35,7 +35,7 @@ def cfg_function(args):
 
             pos = noise_pred_pos - noise_pred_nocond
             neg = noise_pred_neg - noise_pred_nocond
-            perp = ((torch.mul(pos, neg).sum())/(torch.norm(neg)**2)) * neg
+            perp = neg - ((torch.mul(neg, pos).sum())/(torch.norm(pos)**2)) * pos
             perp_neg = perp * neg_scale
             cfg_result = noise_pred_nocond + cond_scale*(pos - perp_neg)
             cfg_result = x - cfg_result
diff --git a/custom_nodes/example_node.py.example b/custom_nodes/example_node.py.example
index 7ce271ec617..f066325930d 100644
--- a/custom_nodes/example_node.py.example
+++ b/custom_nodes/example_node.py.example
@@ -103,6 +103,9 @@ class Example:
     #def IS_CHANGED(s, image, string_field, int_field, float_field, print_to_screen):
     #    return ""
 
+# Set the web directory, any .js file in that directory will be loaded by the frontend as a frontend extension
+# WEB_DIRECTORY = "./somejs"
+
 # A dictionary that contains all nodes you want to export with their names
 # NOTE: names should be globally unique
 NODE_CLASS_MAPPINGS = {
diff --git a/main.py b/main.py
index 69d9bce6cb7..5d07ce2d1fb 100644
--- a/main.py
+++ b/main.py
@@ -193,6 +193,13 @@ def load_extra_path_config(yaml_path):
         folder_paths.set_temp_directory(temp_dir)
     cleanup_temp()
 
+    if args.windows_standalone_build:
+        try:
+            import new_updater
+            new_updater.update_windows_updater()
+        except:
+            pass
+
     loop = asyncio.new_event_loop()
     asyncio.set_event_loop(loop)
     server = server.PromptServer(loop)
diff --git a/new_updater.py b/new_updater.py
new file mode 100644
index 00000000000..a49e0877cb1
--- /dev/null
+++ b/new_updater.py
@@ -0,0 +1,35 @@
+import os
+import shutil
+
+base_path = os.path.dirname(os.path.realpath(__file__))
+
+
+def update_windows_updater():
+    top_path = os.path.dirname(base_path)
+    updater_path = os.path.join(base_path, ".ci/update_windows/update.py")
+    bat_path = os.path.join(base_path, ".ci/update_windows/update_comfyui.bat")
+
+    dest_updater_path = os.path.join(top_path, "update/update.py")
+    dest_bat_path = os.path.join(top_path, "update/update_comfyui.bat")
+    dest_bat_deps_path = os.path.join(top_path, "update/update_comfyui_and_python_dependencies.bat")
+
+    try:
+        with open(dest_bat_path, 'rb') as f:
+            contents = f.read()
+    except:
+        return
+
+    if not contents.startswith(b"..\\python_embeded\\python.exe .\\update.py"):
+        return
+
+    shutil.copy(updater_path, dest_updater_path)
+    try:
+        with open(dest_bat_deps_path, 'rb') as f:
+            contents = f.read()
+            contents = contents.replace(b'..\\python_embeded\\python.exe .\\update.py ..\\ComfyUI\\', b'call update_comfyui.bat nopause')
+        with open(dest_bat_deps_path, 'wb') as f:
+            f.write(contents)
+    except:
+        pass
+    shutil.copy(bat_path, dest_bat_path)
+    print("Updated the windows standalone package updater.")
diff --git a/nodes.py b/nodes.py
index a577c212628..2cfea1a7329 100644
--- a/nodes.py
+++ b/nodes.py
@@ -1003,7 +1003,7 @@ def INPUT_TYPES(s):
 
     def append(self, conditioning_to, clip, gligen_textbox_model, text, width, height, x, y):
         c = []
-        cond, cond_pooled = clip.encode_from_tokens(clip.tokenize(text), return_pooled=True)
+        cond, cond_pooled = clip.encode_from_tokens(clip.tokenize(text), return_pooled="unprojected")
         for t in conditioning_to:
             n = [t[0], t[1].copy()]
             position_params = [(cond_pooled, height // 8, width // 8, y // 8, x // 8)]
@@ -1961,6 +1961,7 @@ def init_custom_nodes():
         "nodes_photomaker.py",
         "nodes_cond.py",
         "nodes_stable_cascade.py",
+        "nodes_differential_diffusion.py",
     ]
 
     for node_file in extras_files:
diff --git a/server.py b/server.py
index dca06f6fc32..c6132cdf91e 100644
--- a/server.py
+++ b/server.py
@@ -539,11 +539,11 @@ def add_routes(self):
 
         for name, dir in nodes.EXTENSION_WEB_DIRS.items():
             self.app.add_routes([
-                web.static('/extensions/' + urllib.parse.quote(name), dir, follow_symlinks=True),
+                web.static('/extensions/' + urllib.parse.quote(name), dir),
             ])
 
         self.app.add_routes([
-            web.static('/', self.web_root, follow_symlinks=True),
+            web.static('/', self.web_root),
         ])
 
     def get_queue_info(self):