Update

8 files changed, 99 insertions, 987 deletions

diff --git a/data/csv.py b/data/csv.py
index fba5d4b..a6cd065 100644
--- a/data/csv.py
+++ b/data/csv.py
@@ -99,14 +99,16 @@ def generate_buckets(
     return buckets, bucket_items, bucket_assignments
 
 
-def collate_fn(dtype: torch.dtype, tokenizer: CLIPTokenizer, with_prior_preservation: bool, examples):
+def collate_fn(dtype: torch.dtype, tokenizer: CLIPTokenizer, with_guidance: bool, with_prior_preservation: bool, examples):
     prompt_ids = [example["prompt_ids"] for example in examples]
     nprompt_ids = [example["nprompt_ids"] for example in examples]
 
     input_ids = [example["instance_prompt_ids"] for example in examples]
     pixel_values = [example["instance_images"] for example in examples]
 
-    if with_prior_preservation:
+    if with_guidance:
+        input_ids += [example["negative_prompt_ids"] for example in examples]
+    elif with_prior_preservation:
         input_ids += [example["class_prompt_ids"] for example in examples]
         pixel_values += [example["class_images"] for example in examples]
 
@@ -133,7 +135,7 @@ class VlpnDataItem(NamedTuple):
     class_image_path: Path
     prompt: list[str]
     cprompt: str
-    nprompt: str
+    nprompt: list[str]
     collection: list[str]
 
 
@@ -163,6 +165,7 @@ class VlpnDataModule():
         data_file: str,
         tokenizer: CLIPTokenizer,
         class_subdir: str = "cls",
+        with_guidance: bool = False,
         num_class_images: int = 1,
         size: int = 768,
         num_buckets: int = 0,
@@ -191,6 +194,7 @@ class VlpnDataModule():
         self.class_root = self.data_root / class_subdir
         self.class_root.mkdir(parents=True, exist_ok=True)
         self.num_class_images = num_class_images
+        self.with_guidance = with_guidance
 
         self.tokenizer = tokenizer
         self.size = size
@@ -228,10 +232,10 @@ class VlpnDataModule():
                     cprompt.format(**prepare_prompt(item["prompt"] if "prompt" in item else "")),
                     expansions
                 )),
-                keywords_to_prompt(prompt_to_keywords(
+                prompt_to_keywords(
                     nprompt.format(**prepare_prompt(item["nprompt"] if "nprompt" in item else "")),
                     expansions
-                )),
+                ),
                 item["collection"].split(", ") if "collection" in item else []
             )
             for item in data
@@ -279,7 +283,7 @@ class VlpnDataModule():
         if self.seed is not None:
             generator = generator.manual_seed(self.seed)
 
-        collate_fn_ = partial(collate_fn, self.dtype, self.tokenizer, self.num_class_images != 0)
+        collate_fn_ = partial(collate_fn, self.dtype, self.tokenizer, self.with_guidance, self.num_class_images != 0)
 
         if valid_set_size == 0:
             data_train, data_val = items, items
@@ -443,11 +447,14 @@ class VlpnDataset(IterableDataset):
             example = {}
 
             example["prompt_ids"] = self.get_input_ids(keywords_to_prompt(item.prompt))
-            example["nprompt_ids"] = self.get_input_ids(item.nprompt)
+            example["nprompt_ids"] = self.get_input_ids(keywords_to_prompt(item.nprompt))
 
             example["instance_prompt_ids"] = self.get_input_ids(
                 keywords_to_prompt(item.prompt, self.dropout, True)
             )
+            example["negative_prompt_ids"] = self.get_input_ids(
+                keywords_to_prompt(item.nprompt, self.dropout, True)
+            )
             example["instance_images"] = image_transforms(get_image(item.instance_image_path))
 
             if self.num_class_images != 0:
diff --git a/pipelines/stable_diffusion/vlpn_stable_diffusion.py b/pipelines/stable_diffusion/vlpn_stable_diffusion.py
index ea2a656..127ca50 100644
--- a/pipelines/stable_diffusion/vlpn_stable_diffusion.py
+++ b/pipelines/stable_diffusion/vlpn_stable_diffusion.py
@@ -307,39 +307,45 @@ class VlpnStableDiffusion(DiffusionPipeline):
 
         return timesteps, num_inference_steps - t_start
 
-    def prepare_image(self, batch_size, width, height, dtype, device, generator=None):
-        return (1.4 * perlin_noise(
+    def prepare_brightness_offset(self, batch_size, height, width, dtype, device, generator=None):
+        offset_image = perlin_noise(
             (batch_size, 1, width, height),
             res=1,
-            octaves=4,
             generator=generator,
             dtype=dtype,
             device=device
-        )).clamp(-1, 1).expand(batch_size, 3, width, height)
+        )
+        offset_latents = self.vae.encode(offset_image).latent_dist.sample(generator=generator)
+        offset_latents = self.vae.config.scaling_factor * offset_latents
+        return offset_latents
 
-    def prepare_latents_from_image(self, init_image, timestep, batch_size, dtype, device, generator=None):
+    def prepare_latents_from_image(self, init_image, timestep, batch_size, brightness_offset, dtype, device, generator=None):
         init_image = init_image.to(device=device, dtype=dtype)
-        init_latents = self.vae.encode(init_image).latent_dist.sample(generator=generator)
-        init_latents = self.vae.config.scaling_factor * init_latents
+        latents = self.vae.encode(init_image).latent_dist.sample(generator=generator)
+        latents = self.vae.config.scaling_factor * latents
 
-        if batch_size % init_latents.shape[0] != 0:
+        if batch_size % latents.shape[0] != 0:
             raise ValueError(
-                f"Cannot duplicate `init_image` of batch size {init_latents.shape[0]} to {batch_size} text prompts."
+                f"Cannot duplicate `init_image` of batch size {latents.shape[0]} to {batch_size} text prompts."
             )
         else:
-            batch_multiplier = batch_size // init_latents.shape[0]
-            init_latents = torch.cat([init_latents] * batch_multiplier, dim=0)
+            batch_multiplier = batch_size // latents.shape[0]
+            latents = torch.cat([latents] * batch_multiplier, dim=0)
 
         # add noise to latents using the timesteps
-        noise = torch.randn(init_latents.shape, generator=generator, device=device, dtype=dtype)
+        noise = torch.randn(latents.shape, generator=generator, device=device, dtype=dtype)
+
+        if brightness_offset != 0:
+            noise += brightness_offset * self.prepare_brightness_offset(
+                batch_size, init_image.shape[3], init_image.shape[2], dtype, device, generator
+            )
 
         # get latents
-        init_latents = self.scheduler.add_noise(init_latents, noise, timestep)
-        latents = init_latents
+        latents = self.scheduler.add_noise(latents, noise, timestep)
 
         return latents
 
-    def prepare_latents(self, batch_size, num_channels_latents, height, width, dtype, device, generator, latents=None):
+    def prepare_latents(self, batch_size, num_channels_latents, height, width, brightness_offset, dtype, device, generator, latents=None):
         shape = (batch_size, num_channels_latents, height // self.vae_scale_factor, width // self.vae_scale_factor)
         if isinstance(generator, list) and len(generator) != batch_size:
             raise ValueError(
@@ -352,6 +358,11 @@ class VlpnStableDiffusion(DiffusionPipeline):
         else:
             latents = latents.to(device)
 
+        if brightness_offset != 0:
+            latents += brightness_offset * self.prepare_brightness_offset(
+                batch_size, height, width, dtype, device, generator
+            )
+
         # scale the initial noise by the standard deviation required by the scheduler
         latents = latents * self.scheduler.init_noise_sigma
         return latents
@@ -395,7 +406,8 @@ class VlpnStableDiffusion(DiffusionPipeline):
         sag_scale: float = 0.75,
         eta: float = 0.0,
         generator: Optional[Union[torch.Generator, List[torch.Generator]]] = None,
-        image: Optional[Union[torch.FloatTensor, PIL.Image.Image, Literal["noise"]]] = None,
+        image: Optional[Union[torch.FloatTensor, PIL.Image.Image]] = None,
+        brightness_offset: Union[float, torch.FloatTensor] = 0,
         output_type: str = "pil",
         return_dict: bool = True,
         callback: Optional[Callable[[int, int, torch.FloatTensor], None]] = None,
@@ -468,7 +480,7 @@ class VlpnStableDiffusion(DiffusionPipeline):
         num_channels_latents = self.unet.in_channels
         do_classifier_free_guidance = guidance_scale > 1.0
         do_self_attention_guidance = sag_scale > 0.0
-        prep_from_image = isinstance(image, PIL.Image.Image) or image == "noise"
+        prep_from_image = isinstance(image, PIL.Image.Image)
 
         # 3. Encode input prompt
         prompt_embeds = self.encode_prompt(
@@ -482,15 +494,6 @@ class VlpnStableDiffusion(DiffusionPipeline):
         # 4. Prepare latent variables
         if isinstance(image, PIL.Image.Image):
             image = preprocess(image)
-        elif image == "noise":
-            image = self.prepare_image(
-                batch_size * num_images_per_prompt,
-                width,
-                height,
-                prompt_embeds.dtype,
-                device,
-                generator
-            )
 
         # 5. Prepare timesteps
         self.scheduler.set_timesteps(num_inference_steps, device=device)
@@ -503,9 +506,10 @@ class VlpnStableDiffusion(DiffusionPipeline):
                 image,
                 latent_timestep,
                 batch_size * num_images_per_prompt,
+                brightness_offset,
                 prompt_embeds.dtype,
                 device,
-                generator
+                generator,
             )
         else:
             latents = self.prepare_latents(
@@ -513,10 +517,11 @@ class VlpnStableDiffusion(DiffusionPipeline):
                 num_channels_latents,
                 height,
                 width,
+                brightness_offset,
                 prompt_embeds.dtype,
                 device,
                 generator,
-                image
+                image,
             )
 
         # 7. Prepare extra step kwargs. TODO: Logic should ideally just be moved out of the pipeline
diff --git a/scripts/convert_diffusers_to_original_stable_diffusion.py b/scripts/convert_diffusers_to_original_stable_diffusion.py
deleted file mode 100644
index 9888f62..0000000
--- a/scripts/convert_diffusers_to_original_stable_diffusion.py
+++ /dev/null
@@ -1,234 +0,0 @@
-# Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint.
-# *Only* converts the UNet, VAE, and Text Encoder.
-# Does not convert optimizer state or any other thing.
-
-import argparse
-import os.path as osp
-
-import torch
-
-
-# =================#
-# UNet Conversion #
-# =================#
-
-unet_conversion_map = [
-    # (stable-diffusion, HF Diffusers)
-    ("time_embed.0.weight", "time_embedding.linear_1.weight"),
-    ("time_embed.0.bias", "time_embedding.linear_1.bias"),
-    ("time_embed.2.weight", "time_embedding.linear_2.weight"),
-    ("time_embed.2.bias", "time_embedding.linear_2.bias"),
-    ("input_blocks.0.0.weight", "conv_in.weight"),
-    ("input_blocks.0.0.bias", "conv_in.bias"),
-    ("out.0.weight", "conv_norm_out.weight"),
-    ("out.0.bias", "conv_norm_out.bias"),
-    ("out.2.weight", "conv_out.weight"),
-    ("out.2.bias", "conv_out.bias"),
-]
-
-unet_conversion_map_resnet = [
-    # (stable-diffusion, HF Diffusers)
-    ("in_layers.0", "norm1"),
-    ("in_layers.2", "conv1"),
-    ("out_layers.0", "norm2"),
-    ("out_layers.3", "conv2"),
-    ("emb_layers.1", "time_emb_proj"),
-    ("skip_connection", "conv_shortcut"),
-]
-
-unet_conversion_map_layer = []
-# hardcoded number of downblocks and resnets/attentions...
-# would need smarter logic for other networks.
-for i in range(4):
-    # loop over downblocks/upblocks
-
-    for j in range(2):
-        # loop over resnets/attentions for downblocks
-        hf_down_res_prefix = f"down_blocks.{i}.resnets.{j}."
-        sd_down_res_prefix = f"input_blocks.{3*i + j + 1}.0."
-        unet_conversion_map_layer.append((sd_down_res_prefix, hf_down_res_prefix))
-
-        if i < 3:
-            # no attention layers in down_blocks.3
-            hf_down_atn_prefix = f"down_blocks.{i}.attentions.{j}."
-            sd_down_atn_prefix = f"input_blocks.{3*i + j + 1}.1."
-            unet_conversion_map_layer.append((sd_down_atn_prefix, hf_down_atn_prefix))
-
-    for j in range(3):
-        # loop over resnets/attentions for upblocks
-        hf_up_res_prefix = f"up_blocks.{i}.resnets.{j}."
-        sd_up_res_prefix = f"output_blocks.{3*i + j}.0."
-        unet_conversion_map_layer.append((sd_up_res_prefix, hf_up_res_prefix))
-
-        if i > 0:
-            # no attention layers in up_blocks.0
-            hf_up_atn_prefix = f"up_blocks.{i}.attentions.{j}."
-            sd_up_atn_prefix = f"output_blocks.{3*i + j}.1."
-            unet_conversion_map_layer.append((sd_up_atn_prefix, hf_up_atn_prefix))
-
-    if i < 3:
-        # no downsample in down_blocks.3
-        hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0.conv."
-        sd_downsample_prefix = f"input_blocks.{3*(i+1)}.0.op."
-        unet_conversion_map_layer.append((sd_downsample_prefix, hf_downsample_prefix))
-
-        # no upsample in up_blocks.3
-        hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
-        sd_upsample_prefix = f"output_blocks.{3*i + 2}.{1 if i == 0 else 2}."
-        unet_conversion_map_layer.append((sd_upsample_prefix, hf_upsample_prefix))
-
-hf_mid_atn_prefix = "mid_block.attentions.0."
-sd_mid_atn_prefix = "middle_block.1."
-unet_conversion_map_layer.append((sd_mid_atn_prefix, hf_mid_atn_prefix))
-
-for j in range(2):
-    hf_mid_res_prefix = f"mid_block.resnets.{j}."
-    sd_mid_res_prefix = f"middle_block.{2*j}."
-    unet_conversion_map_layer.append((sd_mid_res_prefix, hf_mid_res_prefix))
-
-
-def convert_unet_state_dict(unet_state_dict):
-    # buyer beware: this is a *brittle* function,
-    # and correct output requires that all of these pieces interact in
-    # the exact order in which I have arranged them.
-    mapping = {k: k for k in unet_state_dict.keys()}
-    for sd_name, hf_name in unet_conversion_map:
-        mapping[hf_name] = sd_name
-    for k, v in mapping.items():
-        if "resnets" in k:
-            for sd_part, hf_part in unet_conversion_map_resnet:
-                v = v.replace(hf_part, sd_part)
-            mapping[k] = v
-    for k, v in mapping.items():
-        for sd_part, hf_part in unet_conversion_map_layer:
-            v = v.replace(hf_part, sd_part)
-        mapping[k] = v
-    new_state_dict = {v: unet_state_dict[k] for k, v in mapping.items()}
-    return new_state_dict
-
-
-# ================#
-# VAE Conversion #
-# ================#
-
-vae_conversion_map = [
-    # (stable-diffusion, HF Diffusers)
-    ("nin_shortcut", "conv_shortcut"),
-    ("norm_out", "conv_norm_out"),
-    ("mid.attn_1.", "mid_block.attentions.0."),
-]
-
-for i in range(4):
-    # down_blocks have two resnets
-    for j in range(2):
-        hf_down_prefix = f"encoder.down_blocks.{i}.resnets.{j}."
-        sd_down_prefix = f"encoder.down.{i}.block.{j}."
-        vae_conversion_map.append((sd_down_prefix, hf_down_prefix))
-
-    if i < 3:
-        hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0."
-        sd_downsample_prefix = f"down.{i}.downsample."
-        vae_conversion_map.append((sd_downsample_prefix, hf_downsample_prefix))
-
-        hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
-        sd_upsample_prefix = f"up.{3-i}.upsample."
-        vae_conversion_map.append((sd_upsample_prefix, hf_upsample_prefix))
-
-    # up_blocks have three resnets
-    # also, up blocks in hf are numbered in reverse from sd
-    for j in range(3):
-        hf_up_prefix = f"decoder.up_blocks.{i}.resnets.{j}."
-        sd_up_prefix = f"decoder.up.{3-i}.block.{j}."
-        vae_conversion_map.append((sd_up_prefix, hf_up_prefix))
-
-# this part accounts for mid blocks in both the encoder and the decoder
-for i in range(2):
-    hf_mid_res_prefix = f"mid_block.resnets.{i}."
-    sd_mid_res_prefix = f"mid.block_{i+1}."
-    vae_conversion_map.append((sd_mid_res_prefix, hf_mid_res_prefix))
-
-
-vae_conversion_map_attn = [
-    # (stable-diffusion, HF Diffusers)
-    ("norm.", "group_norm."),
-    ("q.", "query."),
-    ("k.", "key."),
-    ("v.", "value."),
-    ("proj_out.", "proj_attn."),
-]
-
-
-def reshape_weight_for_sd(w):
-    # convert HF linear weights to SD conv2d weights
-    return w.reshape(*w.shape, 1, 1)
-
-
-def convert_vae_state_dict(vae_state_dict):
-    mapping = {k: k for k in vae_state_dict.keys()}
-    for k, v in mapping.items():
-        for sd_part, hf_part in vae_conversion_map:
-            v = v.replace(hf_part, sd_part)
-        mapping[k] = v
-    for k, v in mapping.items():
-        if "attentions" in k:
-            for sd_part, hf_part in vae_conversion_map_attn:
-                v = v.replace(hf_part, sd_part)
-            mapping[k] = v
-    new_state_dict = {v: vae_state_dict[k] for k, v in mapping.items()}
-    weights_to_convert = ["q", "k", "v", "proj_out"]
-    for k, v in new_state_dict.items():
-        for weight_name in weights_to_convert:
-            if f"mid.attn_1.{weight_name}.weight" in k:
-                print(f"Reshaping {k} for SD format")
-                new_state_dict[k] = reshape_weight_for_sd(v)
-    return new_state_dict
-
-
-# =========================#
-# Text Encoder Conversion #
-# =========================#
-# pretty much a no-op
-
-
-def convert_text_enc_state_dict(text_enc_dict):
-    return text_enc_dict
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-
-    parser.add_argument("--model_path", default=None, type=str, required=True, help="Path to the model to convert.")
-    parser.add_argument("--checkpoint_path", default=None, type=str, required=True, help="Path to the output model.")
-    parser.add_argument("--half", action="store_true", help="Save weights in half precision.")
-
-    args = parser.parse_args()
-
-    assert args.model_path is not None, "Must provide a model path!"
-
-    assert args.checkpoint_path is not None, "Must provide a checkpoint path!"
-
-    unet_path = osp.join(args.model_path, "unet", "diffusion_pytorch_model.bin")
-    vae_path = osp.join(args.model_path, "vae", "diffusion_pytorch_model.bin")
-    text_enc_path = osp.join(args.model_path, "text_encoder", "pytorch_model.bin")
-
-    # Convert the UNet model
-    unet_state_dict = torch.load(unet_path, map_location="cpu")
-    unet_state_dict = convert_unet_state_dict(unet_state_dict)
-    unet_state_dict = {"model.diffusion_model." + k: v for k, v in unet_state_dict.items()}
-
-    # Convert the VAE model
-    vae_state_dict = torch.load(vae_path, map_location="cpu")
-    vae_state_dict = convert_vae_state_dict(vae_state_dict)
-    vae_state_dict = {"first_stage_model." + k: v for k, v in vae_state_dict.items()}
-
-    # Convert the text encoder model
-    text_enc_dict = torch.load(text_enc_path, map_location="cpu")
-    text_enc_dict = convert_text_enc_state_dict(text_enc_dict)
-    text_enc_dict = {"cond_stage_model.transformer." + k: v for k, v in text_enc_dict.items()}
-
-    # Put together new checkpoint
-    state_dict = {**unet_state_dict, **vae_state_dict, **text_enc_dict}
-    if args.half:
-        state_dict = {k: v.half() for k, v in state_dict.items()}
-    state_dict = {"state_dict": state_dict}
-    torch.save(state_dict, args.checkpoint_path)
diff --git a/scripts/convert_original_stable_diffusion_to_diffusers.py b/scripts/convert_original_stable_diffusion_to_diffusers.py
deleted file mode 100644
index ee7fc33..0000000
--- a/scripts/convert_original_stable_diffusion_to_diffusers.py
+++ /dev/null
@@ -1,690 +0,0 @@
-# coding=utf-8
-# Copyright 2022 The HuggingFace Inc. team.
-#
-# Licensed under the Apache License, Version 2.0 (the "License");
-# you may not use this file except in compliance with the License.
-# You may obtain a copy of the License at
-#
-#     http://www.apache.org/licenses/LICENSE-2.0
-#
-# Unless required by applicable law or agreed to in writing, software
-# distributed under the License is distributed on an "AS IS" BASIS,
-# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
-# See the License for the specific language governing permissions and
-# limitations under the License.
-""" Conversion script for the LDM checkpoints. """
-
-import argparse
-import os
-
-import torch
-
-
-try:
-    from omegaconf import OmegaConf
-except ImportError:
-    raise ImportError(
-        "OmegaConf is required to convert the LDM checkpoints. Please install it with `pip install OmegaConf`."
-    )
-
-from diffusers import (
-    AutoencoderKL,
-    DDIMScheduler,
-    LDMTextToImagePipeline,
-    LMSDiscreteScheduler,
-    PNDMScheduler,
-    StableDiffusionPipeline,
-    UNet2DConditionModel,
-)
-from diffusers.pipelines.latent_diffusion.pipeline_latent_diffusion import LDMBertConfig, LDMBertModel
-from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
-from transformers import AutoFeatureExtractor, BertTokenizerFast, CLIPTextModel, CLIPTokenizer
-
-
-def shave_segments(path, n_shave_prefix_segments=1):
-    """
-    Removes segments. Positive values shave the first segments, negative shave the last segments.
-    """
-    if n_shave_prefix_segments >= 0:
-        return ".".join(path.split(".")[n_shave_prefix_segments:])
-    else:
-        return ".".join(path.split(".")[:n_shave_prefix_segments])
-
-
-def renew_resnet_paths(old_list, n_shave_prefix_segments=0):
-    """
-    Updates paths inside resnets to the new naming scheme (local renaming)
-    """
-    mapping = []
-    for old_item in old_list:
-        new_item = old_item.replace("in_layers.0", "norm1")
-        new_item = new_item.replace("in_layers.2", "conv1")
-
-        new_item = new_item.replace("out_layers.0", "norm2")
-        new_item = new_item.replace("out_layers.3", "conv2")
-
-        new_item = new_item.replace("emb_layers.1", "time_emb_proj")
-        new_item = new_item.replace("skip_connection", "conv_shortcut")
-
-        new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
-
-        mapping.append({"old": old_item, "new": new_item})
-
-    return mapping
-
-
-def renew_vae_resnet_paths(old_list, n_shave_prefix_segments=0):
-    """
-    Updates paths inside resnets to the new naming scheme (local renaming)
-    """
-    mapping = []
-    for old_item in old_list:
-        new_item = old_item
-
-        new_item = new_item.replace("nin_shortcut", "conv_shortcut")
-        new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
-
-        mapping.append({"old": old_item, "new": new_item})
-
-    return mapping
-
-
-def renew_attention_paths(old_list, n_shave_prefix_segments=0):
-    """
-    Updates paths inside attentions to the new naming scheme (local renaming)
-    """
-    mapping = []
-    for old_item in old_list:
-        new_item = old_item
-
-        #         new_item = new_item.replace('norm.weight', 'group_norm.weight')
-        #         new_item = new_item.replace('norm.bias', 'group_norm.bias')
-
-        #         new_item = new_item.replace('proj_out.weight', 'proj_attn.weight')
-        #         new_item = new_item.replace('proj_out.bias', 'proj_attn.bias')
-
-        #         new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
-
-        mapping.append({"old": old_item, "new": new_item})
-
-    return mapping
-
-
-def renew_vae_attention_paths(old_list, n_shave_prefix_segments=0):
-    """
-    Updates paths inside attentions to the new naming scheme (local renaming)
-    """
-    mapping = []
-    for old_item in old_list:
-        new_item = old_item
-
-        new_item = new_item.replace("norm.weight", "group_norm.weight")
-        new_item = new_item.replace("norm.bias", "group_norm.bias")
-
-        new_item = new_item.replace("q.weight", "query.weight")
-        new_item = new_item.replace("q.bias", "query.bias")
-
-        new_item = new_item.replace("k.weight", "key.weight")
-        new_item = new_item.replace("k.bias", "key.bias")
-
-        new_item = new_item.replace("v.weight", "value.weight")
-        new_item = new_item.replace("v.bias", "value.bias")
-
-        new_item = new_item.replace("proj_out.weight", "proj_attn.weight")
-        new_item = new_item.replace("proj_out.bias", "proj_attn.bias")
-
-        new_item = shave_segments(new_item, n_shave_prefix_segments=n_shave_prefix_segments)
-
-        mapping.append({"old": old_item, "new": new_item})
-
-    return mapping
-
-
-def assign_to_checkpoint(
-    paths, checkpoint, old_checkpoint, attention_paths_to_split=None, additional_replacements=None, config=None
-):
-    """
-    This does the final conversion step: take locally converted weights and apply a global renaming
-    to them. It splits attention layers, and takes into account additional replacements
-    that may arise.
-
-    Assigns the weights to the new checkpoint.
-    """
-    assert isinstance(paths, list), "Paths should be a list of dicts containing 'old' and 'new' keys."
-
-    # Splits the attention layers into three variables.
-    if attention_paths_to_split is not None:
-        for path, path_map in attention_paths_to_split.items():
-            old_tensor = old_checkpoint[path]
-            channels = old_tensor.shape[0] // 3
-
-            target_shape = (-1, channels) if len(old_tensor.shape) == 3 else (-1)
-
-            num_heads = old_tensor.shape[0] // config["num_head_channels"] // 3
-
-            old_tensor = old_tensor.reshape((num_heads, 3 * channels // num_heads) + old_tensor.shape[1:])
-            query, key, value = old_tensor.split(channels // num_heads, dim=1)
-
-            checkpoint[path_map["query"]] = query.reshape(target_shape)
-            checkpoint[path_map["key"]] = key.reshape(target_shape)
-            checkpoint[path_map["value"]] = value.reshape(target_shape)
-
-    for path in paths:
-        new_path = path["new"]
-
-        # These have already been assigned
-        if attention_paths_to_split is not None and new_path in attention_paths_to_split:
-            continue
-
-        # Global renaming happens here
-        new_path = new_path.replace("middle_block.0", "mid_block.resnets.0")
-        new_path = new_path.replace("middle_block.1", "mid_block.attentions.0")
-        new_path = new_path.replace("middle_block.2", "mid_block.resnets.1")
-
-        if additional_replacements is not None:
-            for replacement in additional_replacements:
-                new_path = new_path.replace(replacement["old"], replacement["new"])
-
-        # proj_attn.weight has to be converted from conv 1D to linear
-        if "proj_attn.weight" in new_path:
-            checkpoint[new_path] = old_checkpoint[path["old"]][:, :, 0]
-        else:
-            checkpoint[new_path] = old_checkpoint[path["old"]]
-
-
-def conv_attn_to_linear(checkpoint):
-    keys = list(checkpoint.keys())
-    attn_keys = ["query.weight", "key.weight", "value.weight"]
-    for key in keys:
-        if ".".join(key.split(".")[-2:]) in attn_keys:
-            if checkpoint[key].ndim > 2:
-                checkpoint[key] = checkpoint[key][:, :, 0, 0]
-        elif "proj_attn.weight" in key:
-            if checkpoint[key].ndim > 2:
-                checkpoint[key] = checkpoint[key][:, :, 0]
-
-
-def create_unet_diffusers_config(original_config):
-    """
-    Creates a config for the diffusers based on the config of the LDM model.
-    """
-    unet_params = original_config.model.params.unet_config.params
-
-    block_out_channels = [unet_params.model_channels * mult for mult in unet_params.channel_mult]
-
-    down_block_types = []
-    resolution = 1
-    for i in range(len(block_out_channels)):
-        block_type = "CrossAttnDownBlock2D" if resolution in unet_params.attention_resolutions else "DownBlock2D"
-        down_block_types.append(block_type)
-        if i != len(block_out_channels) - 1:
-            resolution *= 2
-
-    up_block_types = []
-    for i in range(len(block_out_channels)):
-        block_type = "CrossAttnUpBlock2D" if resolution in unet_params.attention_resolutions else "UpBlock2D"
-        up_block_types.append(block_type)
-        resolution //= 2
-
-    config = dict(
-        sample_size=unet_params.image_size,
-        in_channels=unet_params.in_channels,
-        out_channels=unet_params.out_channels,
-        down_block_types=tuple(down_block_types),
-        up_block_types=tuple(up_block_types),
-        block_out_channels=tuple(block_out_channels),
-        layers_per_block=unet_params.num_res_blocks,
-        cross_attention_dim=unet_params.context_dim,
-        attention_head_dim=unet_params.num_heads,
-    )
-
-    return config
-
-
-def create_vae_diffusers_config(original_config):
-    """
-    Creates a config for the diffusers based on the config of the LDM model.
-    """
-    vae_params = original_config.model.params.first_stage_config.params.ddconfig
-    _ = original_config.model.params.first_stage_config.params.embed_dim
-
-    block_out_channels = [vae_params.ch * mult for mult in vae_params.ch_mult]
-    down_block_types = ["DownEncoderBlock2D"] * len(block_out_channels)
-    up_block_types = ["UpDecoderBlock2D"] * len(block_out_channels)
-
-    config = dict(
-        sample_size=vae_params.resolution,
-        in_channels=vae_params.in_channels,
-        out_channels=vae_params.out_ch,
-        down_block_types=tuple(down_block_types),
-        up_block_types=tuple(up_block_types),
-        block_out_channels=tuple(block_out_channels),
-        latent_channels=vae_params.z_channels,
-        layers_per_block=vae_params.num_res_blocks,
-    )
-    return config
-
-
-def create_diffusers_schedular(original_config):
-    schedular = DDIMScheduler(
-        num_train_timesteps=original_config.model.params.timesteps,
-        beta_start=original_config.model.params.linear_start,
-        beta_end=original_config.model.params.linear_end,
-        beta_schedule="scaled_linear",
-    )
-    return schedular
-
-
-def create_ldm_bert_config(original_config):
-    bert_params = original_config.model.parms.cond_stage_config.params
-    config = LDMBertConfig(
-        d_model=bert_params.n_embed,
-        encoder_layers=bert_params.n_layer,
-        encoder_ffn_dim=bert_params.n_embed * 4,
-    )
-    return config
-
-
-def convert_ldm_unet_checkpoint(checkpoint, config):
-    """
-    Takes a state dict and a config, and returns a converted checkpoint.
-    """
-
-    # extract state_dict for UNet
-    unet_state_dict = {}
-    unet_key = "model.diffusion_model."
-    keys = list(checkpoint.keys())
-    for key in keys:
-        if key.startswith(unet_key):
-            unet_state_dict[key.replace(unet_key, "")] = checkpoint.pop(key)
-
-    new_checkpoint = {}
-
-    new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]
-    new_checkpoint["time_embedding.linear_1.bias"] = unet_state_dict["time_embed.0.bias"]
-    new_checkpoint["time_embedding.linear_2.weight"] = unet_state_dict["time_embed.2.weight"]
-    new_checkpoint["time_embedding.linear_2.bias"] = unet_state_dict["time_embed.2.bias"]
-
-    new_checkpoint["conv_in.weight"] = unet_state_dict["input_blocks.0.0.weight"]
-    new_checkpoint["conv_in.bias"] = unet_state_dict["input_blocks.0.0.bias"]
-
-    new_checkpoint["conv_norm_out.weight"] = unet_state_dict["out.0.weight"]
-    new_checkpoint["conv_norm_out.bias"] = unet_state_dict["out.0.bias"]
-    new_checkpoint["conv_out.weight"] = unet_state_dict["out.2.weight"]
-    new_checkpoint["conv_out.bias"] = unet_state_dict["out.2.bias"]
-
-    # Retrieves the keys for the input blocks only
-    num_input_blocks = len({".".join(layer.split(".")[:2]) for layer in unet_state_dict if "input_blocks" in layer})
-    input_blocks = {
-        layer_id: [key for key in unet_state_dict if f"input_blocks.{layer_id}" in key]
-        for layer_id in range(num_input_blocks)
-    }
-
-    # Retrieves the keys for the middle blocks only
-    num_middle_blocks = len({".".join(layer.split(".")[:2]) for layer in unet_state_dict if "middle_block" in layer})
-    middle_blocks = {
-        layer_id: [key for key in unet_state_dict if f"middle_block.{layer_id}" in key]
-        for layer_id in range(num_middle_blocks)
-    }
-
-    # Retrieves the keys for the output blocks only
-    num_output_blocks = len({".".join(layer.split(".")[:2]) for layer in unet_state_dict if "output_blocks" in layer})
-    output_blocks = {
-        layer_id: [key for key in unet_state_dict if f"output_blocks.{layer_id}" in key]
-        for layer_id in range(num_output_blocks)
-    }
-
-    for i in range(1, num_input_blocks):
-        block_id = (i - 1) // (config["layers_per_block"] + 1)
-        layer_in_block_id = (i - 1) % (config["layers_per_block"] + 1)
-
-        resnets = [
-            key for key in input_blocks[i] if f"input_blocks.{i}.0" in key and f"input_blocks.{i}.0.op" not in key
-        ]
-        attentions = [key for key in input_blocks[i] if f"input_blocks.{i}.1" in key]
-
-        if f"input_blocks.{i}.0.op.weight" in unet_state_dict:
-            new_checkpoint[f"down_blocks.{block_id}.downsamplers.0.conv.weight"] = unet_state_dict.pop(
-                f"input_blocks.{i}.0.op.weight"
-            )
-            new_checkpoint[f"down_blocks.{block_id}.downsamplers.0.conv.bias"] = unet_state_dict.pop(
-                f"input_blocks.{i}.0.op.bias"
-            )
-
-        paths = renew_resnet_paths(resnets)
-        meta_path = {"old": f"input_blocks.{i}.0", "new": f"down_blocks.{block_id}.resnets.{layer_in_block_id}"}
-        assign_to_checkpoint(
-            paths, new_checkpoint, unet_state_dict, additional_replacements=[meta_path], config=config
-        )
-
-        if len(attentions):
-            paths = renew_attention_paths(attentions)
-            meta_path = {"old": f"input_blocks.{i}.1", "new": f"down_blocks.{block_id}.attentions.{layer_in_block_id}"}
-            assign_to_checkpoint(
-                paths, new_checkpoint, unet_state_dict, additional_replacements=[meta_path], config=config
-            )
-
-    resnet_0 = middle_blocks[0]
-    attentions = middle_blocks[1]
-    resnet_1 = middle_blocks[2]
-
-    resnet_0_paths = renew_resnet_paths(resnet_0)
-    assign_to_checkpoint(resnet_0_paths, new_checkpoint, unet_state_dict, config=config)
-
-    resnet_1_paths = renew_resnet_paths(resnet_1)
-    assign_to_checkpoint(resnet_1_paths, new_checkpoint, unet_state_dict, config=config)
-
-    attentions_paths = renew_attention_paths(attentions)
-    meta_path = {"old": "middle_block.1", "new": "mid_block.attentions.0"}
-    assign_to_checkpoint(
-        attentions_paths, new_checkpoint, unet_state_dict, additional_replacements=[meta_path], config=config
-    )
-
-    for i in range(num_output_blocks):
-        block_id = i // (config["layers_per_block"] + 1)
-        layer_in_block_id = i % (config["layers_per_block"] + 1)
-        output_block_layers = [shave_segments(name, 2) for name in output_blocks[i]]
-        output_block_list = {}
-
-        for layer in output_block_layers:
-            layer_id, layer_name = layer.split(".")[0], shave_segments(layer, 1)
-            if layer_id in output_block_list:
-                output_block_list[layer_id].append(layer_name)
-            else:
-                output_block_list[layer_id] = [layer_name]
-
-        if len(output_block_list) > 1:
-            resnets = [key for key in output_blocks[i] if f"output_blocks.{i}.0" in key]
-            attentions = [key for key in output_blocks[i] if f"output_blocks.{i}.1" in key]
-
-            resnet_0_paths = renew_resnet_paths(resnets)
-            paths = renew_resnet_paths(resnets)
-
-            meta_path = {"old": f"output_blocks.{i}.0", "new": f"up_blocks.{block_id}.resnets.{layer_in_block_id}"}
-            assign_to_checkpoint(
-                paths, new_checkpoint, unet_state_dict, additional_replacements=[meta_path], config=config
-            )
-
-            if ["conv.weight", "conv.bias"] in output_block_list.values():
-                index = list(output_block_list.values()).index(["conv.weight", "conv.bias"])
-                new_checkpoint[f"up_blocks.{block_id}.upsamplers.0.conv.weight"] = unet_state_dict[
-                    f"output_blocks.{i}.{index}.conv.weight"
-                ]
-                new_checkpoint[f"up_blocks.{block_id}.upsamplers.0.conv.bias"] = unet_state_dict[
-                    f"output_blocks.{i}.{index}.conv.bias"
-                ]
-
-                # Clear attentions as they have been attributed above.
-                if len(attentions) == 2:
-                    attentions = []
-
-            if len(attentions):
-                paths = renew_attention_paths(attentions)
-                meta_path = {
-                    "old": f"output_blocks.{i}.1",
-                    "new": f"up_blocks.{block_id}.attentions.{layer_in_block_id}",
-                }
-                assign_to_checkpoint(
-                    paths, new_checkpoint, unet_state_dict, additional_replacements=[meta_path], config=config
-                )
-        else:
-            resnet_0_paths = renew_resnet_paths(output_block_layers, n_shave_prefix_segments=1)
-            for path in resnet_0_paths:
-                old_path = ".".join(["output_blocks", str(i), path["old"]])
-                new_path = ".".join(["up_blocks", str(block_id), "resnets", str(layer_in_block_id), path["new"]])
-
-                new_checkpoint[new_path] = unet_state_dict[old_path]
-
-    return new_checkpoint
-
-
-def convert_ldm_vae_checkpoint(checkpoint, config):
-    # extract state dict for VAE
-    vae_state_dict = {}
-    vae_key = "first_stage_model."
-    keys = list(checkpoint.keys())
-    for key in keys:
-        if key.startswith(vae_key):
-            vae_state_dict[key.replace(vae_key, "")] = checkpoint.get(key)
-
-    new_checkpoint = {}
-
-    new_checkpoint["encoder.conv_in.weight"] = vae_state_dict["encoder.conv_in.weight"]
-    new_checkpoint["encoder.conv_in.bias"] = vae_state_dict["encoder.conv_in.bias"]
-    new_checkpoint["encoder.conv_out.weight"] = vae_state_dict["encoder.conv_out.weight"]
-    new_checkpoint["encoder.conv_out.bias"] = vae_state_dict["encoder.conv_out.bias"]
-    new_checkpoint["encoder.conv_norm_out.weight"] = vae_state_dict["encoder.norm_out.weight"]
-    new_checkpoint["encoder.conv_norm_out.bias"] = vae_state_dict["encoder.norm_out.bias"]
-
-    new_checkpoint["decoder.conv_in.weight"] = vae_state_dict["decoder.conv_in.weight"]
-    new_checkpoint["decoder.conv_in.bias"] = vae_state_dict["decoder.conv_in.bias"]
-    new_checkpoint["decoder.conv_out.weight"] = vae_state_dict["decoder.conv_out.weight"]
-    new_checkpoint["decoder.conv_out.bias"] = vae_state_dict["decoder.conv_out.bias"]
-    new_checkpoint["decoder.conv_norm_out.weight"] = vae_state_dict["decoder.norm_out.weight"]
-    new_checkpoint["decoder.conv_norm_out.bias"] = vae_state_dict["decoder.norm_out.bias"]
-
-    new_checkpoint["quant_conv.weight"] = vae_state_dict["quant_conv.weight"]
-    new_checkpoint["quant_conv.bias"] = vae_state_dict["quant_conv.bias"]
-    new_checkpoint["post_quant_conv.weight"] = vae_state_dict["post_quant_conv.weight"]
-    new_checkpoint["post_quant_conv.bias"] = vae_state_dict["post_quant_conv.bias"]
-
-    # Retrieves the keys for the encoder down blocks only
-    num_down_blocks = len({".".join(layer.split(".")[:3]) for layer in vae_state_dict if "encoder.down" in layer})
-    down_blocks = {
-        layer_id: [key for key in vae_state_dict if f"down.{layer_id}" in key] for layer_id in range(num_down_blocks)
-    }
-
-    # Retrieves the keys for the decoder up blocks only
-    num_up_blocks = len({".".join(layer.split(".")[:3]) for layer in vae_state_dict if "decoder.up" in layer})
-    up_blocks = {
-        layer_id: [key for key in vae_state_dict if f"up.{layer_id}" in key] for layer_id in range(num_up_blocks)
-    }
-
-    for i in range(num_down_blocks):
-        resnets = [key for key in down_blocks[i] if f"down.{i}" in key and f"down.{i}.downsample" not in key]
-
-        if f"encoder.down.{i}.downsample.conv.weight" in vae_state_dict:
-            new_checkpoint[f"encoder.down_blocks.{i}.downsamplers.0.conv.weight"] = vae_state_dict.pop(
-                f"encoder.down.{i}.downsample.conv.weight"
-            )
-            new_checkpoint[f"encoder.down_blocks.{i}.downsamplers.0.conv.bias"] = vae_state_dict.pop(
-                f"encoder.down.{i}.downsample.conv.bias"
-            )
-
-        paths = renew_vae_resnet_paths(resnets)
-        meta_path = {"old": f"down.{i}.block", "new": f"down_blocks.{i}.resnets"}
-        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-
-    mid_resnets = [key for key in vae_state_dict if "encoder.mid.block" in key]
-    num_mid_res_blocks = 2
-    for i in range(1, num_mid_res_blocks + 1):
-        resnets = [key for key in mid_resnets if f"encoder.mid.block_{i}" in key]
-
-        paths = renew_vae_resnet_paths(resnets)
-        meta_path = {"old": f"mid.block_{i}", "new": f"mid_block.resnets.{i - 1}"}
-        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-
-    mid_attentions = [key for key in vae_state_dict if "encoder.mid.attn" in key]
-    paths = renew_vae_attention_paths(mid_attentions)
-    meta_path = {"old": "mid.attn_1", "new": "mid_block.attentions.0"}
-    assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-    conv_attn_to_linear(new_checkpoint)
-
-    for i in range(num_up_blocks):
-        block_id = num_up_blocks - 1 - i
-        resnets = [
-            key for key in up_blocks[block_id] if f"up.{block_id}" in key and f"up.{block_id}.upsample" not in key
-        ]
-
-        if f"decoder.up.{block_id}.upsample.conv.weight" in vae_state_dict:
-            new_checkpoint[f"decoder.up_blocks.{i}.upsamplers.0.conv.weight"] = vae_state_dict[
-                f"decoder.up.{block_id}.upsample.conv.weight"
-            ]
-            new_checkpoint[f"decoder.up_blocks.{i}.upsamplers.0.conv.bias"] = vae_state_dict[
-                f"decoder.up.{block_id}.upsample.conv.bias"
-            ]
-
-        paths = renew_vae_resnet_paths(resnets)
-        meta_path = {"old": f"up.{block_id}.block", "new": f"up_blocks.{i}.resnets"}
-        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-
-    mid_resnets = [key for key in vae_state_dict if "decoder.mid.block" in key]
-    num_mid_res_blocks = 2
-    for i in range(1, num_mid_res_blocks + 1):
-        resnets = [key for key in mid_resnets if f"decoder.mid.block_{i}" in key]
-
-        paths = renew_vae_resnet_paths(resnets)
-        meta_path = {"old": f"mid.block_{i}", "new": f"mid_block.resnets.{i - 1}"}
-        assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-
-    mid_attentions = [key for key in vae_state_dict if "decoder.mid.attn" in key]
-    paths = renew_vae_attention_paths(mid_attentions)
-    meta_path = {"old": "mid.attn_1", "new": "mid_block.attentions.0"}
-    assign_to_checkpoint(paths, new_checkpoint, vae_state_dict, additional_replacements=[meta_path], config=config)
-    conv_attn_to_linear(new_checkpoint)
-    return new_checkpoint
-
-
-def convert_ldm_bert_checkpoint(checkpoint, config):
-    def _copy_attn_layer(hf_attn_layer, pt_attn_layer):
-        hf_attn_layer.q_proj.weight.data = pt_attn_layer.to_q.weight
-        hf_attn_layer.k_proj.weight.data = pt_attn_layer.to_k.weight
-        hf_attn_layer.v_proj.weight.data = pt_attn_layer.to_v.weight
-
-        hf_attn_layer.out_proj.weight = pt_attn_layer.to_out.weight
-        hf_attn_layer.out_proj.bias = pt_attn_layer.to_out.bias
-
-    def _copy_linear(hf_linear, pt_linear):
-        hf_linear.weight = pt_linear.weight
-        hf_linear.bias = pt_linear.bias
-
-    def _copy_layer(hf_layer, pt_layer):
-        # copy layer norms
-        _copy_linear(hf_layer.self_attn_layer_norm, pt_layer[0][0])
-        _copy_linear(hf_layer.final_layer_norm, pt_layer[1][0])
-
-        # copy attn
-        _copy_attn_layer(hf_layer.self_attn, pt_layer[0][1])
-
-        # copy MLP
-        pt_mlp = pt_layer[1][1]
-        _copy_linear(hf_layer.fc1, pt_mlp.net[0][0])
-        _copy_linear(hf_layer.fc2, pt_mlp.net[2])
-
-    def _copy_layers(hf_layers, pt_layers):
-        for i, hf_layer in enumerate(hf_layers):
-            if i != 0:
-                i += i
-            pt_layer = pt_layers[i : i + 2]
-            _copy_layer(hf_layer, pt_layer)
-
-    hf_model = LDMBertModel(config).eval()
-
-    # copy  embeds
-    hf_model.model.embed_tokens.weight = checkpoint.transformer.token_emb.weight
-    hf_model.model.embed_positions.weight.data = checkpoint.transformer.pos_emb.emb.weight
-
-    # copy layer norm
-    _copy_linear(hf_model.model.layer_norm, checkpoint.transformer.norm)
-
-    # copy hidden layers
-    _copy_layers(hf_model.model.layers, checkpoint.transformer.attn_layers.layers)
-
-    _copy_linear(hf_model.to_logits, checkpoint.transformer.to_logits)
-
-    return hf_model
-
-
-if __name__ == "__main__":
-    parser = argparse.ArgumentParser()
-
-    parser.add_argument(
-        "--checkpoint_path", default=None, type=str, required=True, help="Path to the checkpoint to convert."
-    )
-    # !wget https://raw.githubusercontent.com/CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml
-    parser.add_argument(
-        "--original_config_file",
-        default=None,
-        type=str,
-        help="The YAML config file corresponding to the original architecture.",
-    )
-    parser.add_argument(
-        "--scheduler_type",
-        default="pndm",
-        type=str,
-        help="Type of scheduler to use. Should be one of ['pndm', 'lms', 'ddim']",
-    )
-    parser.add_argument("--dump_path", default=None, type=str, required=True, help="Path to the output model.")
-
-    args = parser.parse_args()
-
-    if args.original_config_file is None:
-        os.system(
-            "wget https://raw.githubusercontent.com/CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml"
-        )
-        args.original_config_file = "./v1-inference.yaml"
-
-    original_config = OmegaConf.load(args.original_config_file)
-    checkpoint = torch.load(args.checkpoint_path)["state_dict"]
-
-    num_train_timesteps = original_config.model.params.timesteps
-    beta_start = original_config.model.params.linear_start
-    beta_end = original_config.model.params.linear_end
-    if args.scheduler_type == "pndm":
-        scheduler = PNDMScheduler(
-            beta_end=beta_end,
-            beta_schedule="scaled_linear",
-            beta_start=beta_start,
-            num_train_timesteps=num_train_timesteps,
-            skip_prk_steps=True,
-        )
-    elif args.scheduler_type == "lms":
-        scheduler = LMSDiscreteScheduler(beta_start=beta_start, beta_end=beta_end, beta_schedule="scaled_linear")
-    elif args.scheduler_type == "ddim":
-        scheduler = DDIMScheduler(
-            beta_start=beta_start,
-            beta_end=beta_end,
-            beta_schedule="scaled_linear",
-            clip_sample=False,
-            set_alpha_to_one=False,
-        )
-    else:
-        raise ValueError(f"Scheduler of type {args.scheduler_type} doesn't exist!")
-
-    # Convert the UNet2DConditionModel model.
-    unet_config = create_unet_diffusers_config(original_config)
-    converted_unet_checkpoint = convert_ldm_unet_checkpoint(checkpoint, unet_config)
-
-    unet = UNet2DConditionModel(**unet_config)
-    unet.load_state_dict(converted_unet_checkpoint)
-
-    # Convert the VAE model.
-    vae_config = create_vae_diffusers_config(original_config)
-    converted_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
-
-    vae = AutoencoderKL(**vae_config)
-    vae.load_state_dict(converted_vae_checkpoint)
-
-    # Convert the text model.
-    text_model_type = original_config.model.params.cond_stage_config.target.split(".")[-1]
-    if text_model_type == "FrozenCLIPEmbedder":
-        text_model = CLIPTextModel.from_pretrained("openai/clip-vit-large-patch14")
-        tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
-        safety_checker = StableDiffusionSafetyChecker.from_pretrained("CompVis/stable-diffusion-safety-checker")
-        feature_extractor = AutoFeatureExtractor.from_pretrained("CompVis/stable-diffusion-safety-checker")
-        pipe = StableDiffusionPipeline(
-            vae=vae,
-            text_encoder=text_model,
-            tokenizer=tokenizer,
-            unet=unet,
-            scheduler=scheduler,
-            safety_checker=safety_checker,
-            feature_extractor=feature_extractor,
-        )
-    else:
-        text_config = create_ldm_bert_config(original_config)
-        text_model = convert_ldm_bert_checkpoint(checkpoint, text_config)
-        tokenizer = BertTokenizerFast.from_pretrained("bert-base-uncased")
-        pipe = LDMTextToImagePipeline(vqvae=vae, bert=text_model, tokenizer=tokenizer, unet=unet, scheduler=scheduler)
-
-    pipe.save_pretrained(args.dump_path)
diff --git a/train_dreambooth.py b/train_dreambooth.py
index 1b8a3d2..7a33bca 100644
--- a/train_dreambooth.py
+++ b/train_dreambooth.py
@@ -110,7 +110,7 @@ def parse_args():
     parser.add_argument(
         "--tag_dropout",
         type=float,
-        default=0.1,
+        default=0,
         help="Tag dropout probability.",
     )
     parser.add_argument(
@@ -131,6 +131,11 @@ def parse_args():
         help='Vector shuffling algorithm. Choose between ["all", "trailing", "leading", "between", "auto", "off"]',
     )
     parser.add_argument(
+        "--guidance_scale",
+        type=float,
+        default=0,
+    )
+    parser.add_argument(
         "--num_class_images",
         type=int,
         default=0,
@@ -178,7 +183,7 @@ def parse_args():
     parser.add_argument(
         "--offset_noise_strength",
         type=float,
-        default=0.15,
+        default=0,
         help="Perlin offset noise strength.",
     )
     parser.add_argument(
@@ -557,8 +562,8 @@ def main():
         vae=vae,
         noise_scheduler=noise_scheduler,
         dtype=weight_dtype,
-        with_prior_preservation=args.num_class_images != 0,
-        prior_loss_weight=args.prior_loss_weight,
+        guidance_scale=args.guidance_scale,
+        prior_loss_weight=args.prior_loss_weight if args.num_class_images != 0 else 0,
         no_val=args.valid_set_size == 0,
     )
 
@@ -570,6 +575,7 @@ def main():
         batch_size=args.train_batch_size,
         tokenizer=tokenizer,
         class_subdir=args.class_image_dir,
+        with_guidance=args.guidance_scale != 0,
         num_class_images=args.num_class_images,
         size=args.resolution,
         num_buckets=args.num_buckets,
diff --git a/train_lora.py b/train_lora.py
index b16a99b..684d0cc 100644
--- a/train_lora.py
+++ b/train_lora.py
@@ -88,7 +88,7 @@ def parse_args():
     parser.add_argument(
         "--num_buckets",
         type=int,
-        default=0,
+        default=2,
         help="Number of aspect ratio buckets in either direction.",
     )
     parser.add_argument(
@@ -111,7 +111,7 @@ def parse_args():
     parser.add_argument(
         "--tag_dropout",
         type=float,
-        default=0.1,
+        default=0,
         help="Tag dropout probability.",
     )
     parser.add_argument(
@@ -120,6 +120,11 @@ def parse_args():
         help="Shuffle tags.",
     )
     parser.add_argument(
+        "--guidance_scale",
+        type=float,
+        default=0,
+    )
+    parser.add_argument(
         "--num_class_images",
         type=int,
         default=0,
@@ -167,7 +172,7 @@ def parse_args():
     parser.add_argument(
         "--offset_noise_strength",
         type=float,
-        default=0.15,
+        default=0,
         help="Perlin offset noise strength.",
     )
     parser.add_argument(
@@ -589,8 +594,8 @@ def main():
         vae=vae,
         noise_scheduler=noise_scheduler,
         dtype=weight_dtype,
-        with_prior_preservation=args.num_class_images != 0,
-        prior_loss_weight=args.prior_loss_weight,
+        guidance_scale=args.guidance_scale,
+        prior_loss_weight=args.prior_loss_weight if args.num_class_images != 0 else 0,
         no_val=args.valid_set_size == 0,
     )
 
@@ -602,6 +607,7 @@ def main():
         batch_size=args.train_batch_size,
         tokenizer=tokenizer,
         class_subdir=args.class_image_dir,
+        with_guidance=args.guidance_scale != 0,
         num_class_images=args.num_class_images,
         size=args.resolution,
         num_buckets=args.num_buckets,
diff --git a/train_ti.py b/train_ti.py
index bbc5524..83ad46d 100644
--- a/train_ti.py
+++ b/train_ti.py
@@ -91,6 +91,11 @@ def parse_args():
         action="store_true",
     )
     parser.add_argument(
+        "--guidance_scale",
+        type=float,
+        default=0,
+    )
+    parser.add_argument(
         "--num_class_images",
         type=int,
         default=0,
@@ -167,7 +172,7 @@ def parse_args():
     parser.add_argument(
         "--tag_dropout",
         type=float,
-        default=0.1,
+        default=0,
         help="Tag dropout probability.",
     )
     parser.add_argument(
@@ -190,7 +195,7 @@ def parse_args():
     parser.add_argument(
         "--offset_noise_strength",
         type=float,
-        default=0.15,
+        default=0,
         help="Perlin offset noise strength.",
     )
     parser.add_argument(
@@ -651,8 +656,8 @@ def main():
         noise_scheduler=noise_scheduler,
         dtype=weight_dtype,
         seed=args.seed,
-        with_prior_preservation=args.num_class_images != 0,
-        prior_loss_weight=args.prior_loss_weight,
+        guidance_scale=args.guidance_scale,
+        prior_loss_weight=args.prior_loss_weight if args.num_class_images != 0 else 0,
         no_val=args.valid_set_size == 0,
         strategy=textual_inversion_strategy,
         num_train_epochs=args.num_train_epochs,
@@ -705,6 +710,7 @@ def main():
             batch_size=args.train_batch_size,
             tokenizer=tokenizer,
             class_subdir=args.class_image_dir,
+            with_guidance=args.guidance_scale != 0,
             num_class_images=args.num_class_images,
             size=args.resolution,
             num_buckets=args.num_buckets,
diff --git a/training/functional.py b/training/functional.py
index 87bb339..d285366 100644
--- a/training/functional.py
+++ b/training/functional.py
@@ -274,7 +274,7 @@ def loss_step(
     noise_scheduler: SchedulerMixin,
     unet: UNet2DConditionModel,
     text_encoder: CLIPTextModel,
-    with_prior_preservation: bool,
+    guidance_scale: float,
     prior_loss_weight: float,
     seed: int,
     offset_noise_strength: float,
@@ -283,13 +283,13 @@ def loss_step(
     eval: bool = False,
     min_snr_gamma: int = 5,
 ):
-    # Convert images to latent space
-    latents = vae.encode(batch["pixel_values"]).latent_dist.sample()
-    latents = latents * vae.config.scaling_factor
-
-    bsz = latents.shape[0]
+    images = batch["pixel_values"]
+    generator = torch.Generator(device=images.device).manual_seed(seed + step) if eval else None
+    bsz = images.shape[0]
 
-    generator = torch.Generator(device=latents.device).manual_seed(seed + step) if eval else None
+    # Convert images to latent space
+    latents = vae.encode(images).latent_dist.sample(generator=generator)
+    latents *= vae.config.scaling_factor
 
     # Sample noise that we'll add to the latents
     noise = torch.randn(
@@ -301,13 +301,13 @@ def loss_step(
     )
 
     if offset_noise_strength != 0:
-        noise += offset_noise_strength * perlin_noise(
-            latents.shape,
-            res=1,
+        offset_noise = torch.randn(
+            (latents.shape[0], latents.shape[1], 1, 1),
             dtype=latents.dtype,
             device=latents.device,
             generator=generator
-        )
+        ).expand(noise.shape)
+        noise += offset_noise_strength * offset_noise
 
     # Sample a random timestep for each image
     timesteps = torch.randint(
@@ -343,7 +343,13 @@ def loss_step(
     else:
         raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
 
-    if with_prior_preservation:
+    if guidance_scale != 0:
+        # Chunk the noise and model_pred into two parts and compute the loss on each part separately.
+        model_pred_uncond, model_pred_text = torch.chunk(model_pred, 2, dim=0)
+        model_pred = model_pred_uncond + guidance_scale * (model_pred_text - model_pred_uncond)
+
+        loss = F.mse_loss(model_pred.float(), target.float(), reduction="none")
+    elif prior_loss_weight != 0:
         # Chunk the noise and model_pred into two parts and compute the loss on each part separately.
         model_pred, model_pred_prior = torch.chunk(model_pred, 2, dim=0)
         target, target_prior = torch.chunk(target, 2, dim=0)
@@ -607,9 +613,9 @@ def train(
     checkpoint_frequency: int = 50,
     milestone_checkpoints: bool = True,
     global_step_offset: int = 0,
-    with_prior_preservation: bool = False,
+    guidance_scale: float = 0.0,
     prior_loss_weight: float = 1.0,
-    offset_noise_strength: float = 0.1,
+    offset_noise_strength: float = 0.15,
     **kwargs,
 ):
     text_encoder, unet, optimizer, train_dataloader, val_dataloader, lr_scheduler, extra = strategy.prepare(
@@ -638,7 +644,7 @@ def train(
         noise_scheduler,
         unet,
         text_encoder,
-        with_prior_preservation,
+        guidance_scale,
         prior_loss_weight,
         seed,
         offset_noise_strength,