Convert original Stable Diffusion checkpoints and safetensors to diffusers

$README.md

Stable Diffusion model conversion script

Convert from the 'original implementation' to Huggingface diffusers. Both .safetensors and .ckpt checkpoints are supported.

The script is adapted from the diffusers conversion script: https://github.com/huggingface/diffusers/blob/main/scripts/convert_original_stable_diffusion_to_diffusers.py

Usage:

python enhanced_convert_original_stable_diffusion_to_diffusers.py --checkpoint_path "<input_model_filepath>" --dump_path "<output_directory_path>" --extract_ema

For inference use, extracting with --extract_ema is usually recommended. This will be ignored if the source model does not contain both EMA and non-EMA weights.

A workaround for the following error is implemented:

RuntimeError: Error(s) in loading state_dict for UNet2DConditionModel:
        Missing key(s) in state_dict: "up_blocks.0.upsamplers.0.conv.weight", "up_blocks.0.upsamplers.0.conv.bias", "up_blocks.1.upsamplers.0.conv.weight", "up_blocks.1.upsamplers.0.conv.bias", "up_blocks.2.upsamplers.0.conv.weight", "up_blocks.2.upsamplers.0.conv.bias".
        Unexpected key(s) in state_dict: "up_blocks.0.attentions.2.conv.bias", "up_blocks.0.attentions.2.conv.weight".

This has been found to occur when converting some models, especially models that were distributed in a safetensors format.

To extract and re-add missing keys of the state dictionary from the original model, the key map from https://github.com/ratwithacompiler/diffusers_stablediff_conversion/blob/main/convert_diffusers_to_sd.py was used.

In case of problems with other UNET keys, lines 911 and 913 of the script can be extended:

• missing_diffusers_keys_unet will retrieve and amend any diffusers UNET keys (for keys reported as Missing key(s) in state_dict).
• unwanted_diffusers_keys_unet will remove any unwanted keys (for keys reported as Unexpected key(s) in state_dict).

enhanced_convert_original_stable_diffusion_to_diffusers.py

# 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. """

# Modified with a workaround for safetensors support and UNET state_dict problems.

import argparse
import os
import re

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,
    #DPMSolverMultistepScheduler,
    #EulerAncestralDiscreteScheduler,
    EulerDiscreteScheduler,
    #HeunDiscreteScheduler,
    LDMTextToImagePipeline,
    LMSDiscreteScheduler,
    PNDMScheduler,
    StableDiffusionPipeline,
    UNet2DConditionModel,
)
from diffusers.pipelines.latent_diffusion.pipeline_latent_diffusion import LDMBertConfig, LDMBertModel
#from diffusers.pipelines.paint_by_example import PaintByExampleImageEncoder, PaintByExamplePipeline
from diffusers.pipelines.stable_diffusion import StableDiffusionSafetyChecker
from transformers import AutoFeatureExtractor, BertTokenizerFast, CLIPTextModel, CLIPTokenizer, CLIPVisionConfig

from safetensors.torch import load_file

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, image_size: int):
    """
    Creates a config for the diffusers based on the config of the LDM model.
    """
    unet_params = original_config.model.params.unet_config.params
    vae_params = original_config.model.params.first_stage_config.params.ddconfig

    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

    vae_scale_factor = 2 ** (len(vae_params.ch_mult) - 1)

    head_dim = unet_params.num_heads if "num_heads" in unet_params else None
    use_linear_projection = (
        unet_params.use_linear_in_transformer if "use_linear_in_transformer" in unet_params else False
    )
    if use_linear_projection:
        # stable diffusion 2-base-512 and 2-768
        if head_dim is None:
            head_dim = [5, 10, 20, 20]

    config = dict(
        sample_size=image_size // vae_scale_factor,
        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=head_dim,
        use_linear_projection=use_linear_projection,
    )

    return config


def create_vae_diffusers_config(original_config, image_size: int):
    """
    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=image_size,
        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, path=None, extract_ema=False):
    """
    Takes a state dict and a config, and returns a converted checkpoint.
    """

    # extract state_dict for UNet
    unet_state_dict = {}
    keys = list(checkpoint.keys())

    unet_key = "model.diffusion_model."
    # at least a 100 parameters have to start with `model_ema` in order for the checkpoint to be EMA
    if sum(k.startswith("model_ema") for k in keys) > 100:
        print(f"Checkpoint {path} has both EMA and non-EMA weights.")
        if extract_ema:
            print(
                "In this conversion only the EMA weights are extracted. If you want to instead extract the non-EMA"
                " weights (useful to continue fine-tuning), please make sure to remove the `--extract_ema` flag."
            )
            for key in keys:
                if key.startswith("model.diffusion_model"):
                    flat_ema_key = "model_ema." + "".join(key.split(".")[1:])
                    unet_state_dict[key.replace(unet_key, "")] = checkpoint.pop(flat_ema_key)
        else:
            print(
                "In this conversion only the non-EMA weights are extracted. If you want to instead extract the EMA"
                " weights (usually better for inference), please make sure to add the `--extract_ema` flag."
            )

    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


def convert_ldm_clip_checkpoint(checkpoint):
    text_model = CLIPTextModel.from_pretrained("openai/clip-vit-large-patch14")

    keys = list(checkpoint.keys())

    text_model_dict = {}

    for key in keys:
        if key.startswith("cond_stage_model.transformer"):
            text_model_dict[key[len("cond_stage_model.transformer.") :]] = checkpoint[key]

    text_model.load_state_dict(text_model_dict)

    return text_model


textenc_conversion_lst = [
    ("cond_stage_model.model.positional_embedding", "text_model.embeddings.position_embedding.weight"),
    ("cond_stage_model.model.token_embedding.weight", "text_model.embeddings.token_embedding.weight"),
    ("cond_stage_model.model.ln_final.weight", "text_model.final_layer_norm.weight"),
    ("cond_stage_model.model.ln_final.bias", "text_model.final_layer_norm.bias"),
]
textenc_conversion_map = {x[0]: x[1] for x in textenc_conversion_lst}

textenc_transformer_conversion_lst = [
    # (stable-diffusion, HF Diffusers)
    ("resblocks.", "text_model.encoder.layers."),
    ("ln_1", "layer_norm1"),
    ("ln_2", "layer_norm2"),
    (".c_fc.", ".fc1."),
    (".c_proj.", ".fc2."),
    (".attn", ".self_attn"),
    ("ln_final.", "transformer.text_model.final_layer_norm."),
    ("token_embedding.weight", "transformer.text_model.embeddings.token_embedding.weight"),
    ("positional_embedding", "transformer.text_model.embeddings.position_embedding.weight"),
]
protected = {re.escape(x[0]): x[1] for x in textenc_transformer_conversion_lst}
textenc_pattern = re.compile("|".join(protected.keys()))


def convert_paint_by_example_checkpoint(checkpoint):
    config = CLIPVisionConfig.from_pretrained("openai/clip-vit-large-patch14")
    model = PaintByExampleImageEncoder(config)

    keys = list(checkpoint.keys())

    text_model_dict = {}

    for key in keys:
        if key.startswith("cond_stage_model.transformer"):
            text_model_dict[key[len("cond_stage_model.transformer.") :]] = checkpoint[key]

    # load clip vision
    model.model.load_state_dict(text_model_dict)

    # load mapper
    keys_mapper = {
        k[len("cond_stage_model.mapper.res") :]: v
        for k, v in checkpoint.items()
        if k.startswith("cond_stage_model.mapper")
    }

    MAPPING = {
        "attn.c_qkv": ["attn1.to_q", "attn1.to_k", "attn1.to_v"],
        "attn.c_proj": ["attn1.to_out.0"],
        "ln_1": ["norm1"],
        "ln_2": ["norm3"],
        "mlp.c_fc": ["ff.net.0.proj"],
        "mlp.c_proj": ["ff.net.2"],
    }

    mapped_weights = {}
    for key, value in keys_mapper.items():
        prefix = key[: len("blocks.i")]
        suffix = key.split(prefix)[-1].split(".")[-1]
        name = key.split(prefix)[-1].split(suffix)[0][1:-1]
        mapped_names = MAPPING[name]

        num_splits = len(mapped_names)
        for i, mapped_name in enumerate(mapped_names):
            new_name = ".".join([prefix, mapped_name, suffix])
            shape = value.shape[0] // num_splits
            mapped_weights[new_name] = value[i * shape : (i + 1) * shape]

    model.mapper.load_state_dict(mapped_weights)

    # load final layer norm
    model.final_layer_norm.load_state_dict(
        {
            "bias": checkpoint["cond_stage_model.final_ln.bias"],
            "weight": checkpoint["cond_stage_model.final_ln.weight"],
        }
    )

    # load final proj
    model.proj_out.load_state_dict(
        {
            "bias": checkpoint["proj_out.bias"],
            "weight": checkpoint["proj_out.weight"],
        }
    )

    # load uncond vector
    model.uncond_vector.data = torch.nn.Parameter(checkpoint["learnable_vector"])
    return model


def convert_open_clip_checkpoint(checkpoint):
    text_model = CLIPTextModel.from_pretrained("stabilityai/stable-diffusion-2", subfolder="text_encoder")

    keys = list(checkpoint.keys())

    text_model_dict = {}

    d_model = int(checkpoint["cond_stage_model.model.text_projection"].shape[0])

    text_model_dict["text_model.embeddings.position_ids"] = text_model.text_model.embeddings.get_buffer("position_ids")

    for key in keys:
        if "resblocks.23" in key:  # Diffusers drops the final layer and only uses the penultimate layer
            continue
        if key in textenc_conversion_map:
            text_model_dict[textenc_conversion_map[key]] = checkpoint[key]
        if key.startswith("cond_stage_model.model.transformer."):
            new_key = key[len("cond_stage_model.model.transformer.") :]
            if new_key.endswith(".in_proj_weight"):
                new_key = new_key[: -len(".in_proj_weight")]
                new_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], new_key)
                text_model_dict[new_key + ".q_proj.weight"] = checkpoint[key][:d_model, :]
                text_model_dict[new_key + ".k_proj.weight"] = checkpoint[key][d_model : d_model * 2, :]
                text_model_dict[new_key + ".v_proj.weight"] = checkpoint[key][d_model * 2 :, :]
            elif new_key.endswith(".in_proj_bias"):
                new_key = new_key[: -len(".in_proj_bias")]
                new_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], new_key)
                text_model_dict[new_key + ".q_proj.bias"] = checkpoint[key][:d_model]
                text_model_dict[new_key + ".k_proj.bias"] = checkpoint[key][d_model : d_model * 2]
                text_model_dict[new_key + ".v_proj.bias"] = checkpoint[key][d_model * 2 :]
            else:
                new_key = textenc_pattern.sub(lambda m: protected[re.escape(m.group(0))], new_key)

                text_model_dict[new_key] = checkpoint[key]

    text_model.load_state_dict(text_model_dict)

    return text_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(
        "--num_in_channels",
        default=None,
        type=int,
        help="The number of input channels. If `None` number of input channels will be automatically inferred.",
    )
    parser.add_argument(
        "--scheduler_type",
        default="pndm",
        type=str,
        help="Type of scheduler to use. Should be one of ['pndm', 'lms', 'ddim', 'euler', 'euler-ancestral', 'dpm']",
    )
    parser.add_argument(
        "--pipeline_type",
        default=None,
        type=str,
        help="The pipeline type. If `None` pipeline will be automatically inferred.",
    )
    parser.add_argument(
        "--image_size",
        default=None,
        type=int,
        help=(
            "The image size that the model was trained on. Use 512 for Stable Diffusion v1.X and Stable Siffusion v2"
            " Base. Use 768 for Stable Diffusion v2."
        ),
    )
    parser.add_argument(
        "--prediction_type",
        default=None,
        type=str,
        help=(
            "The prediction type that the model was trained on. Use 'epsilon' for Stable Diffusion v1.X and Stable"
            " Siffusion v2 Base. Use 'v-prediction' for Stable Diffusion v2."
        ),
    )
    parser.add_argument(
        "--extract_ema",
        action="store_true",
        help=(
            "Only relevant for checkpoints that have both EMA and non-EMA weights. Whether to extract the EMA weights"
            " or not. Defaults to `False`. Add `--extract_ema` to extract the EMA weights. EMA weights usually yield"
            " higher quality images for inference. Non-EMA weights are usually better to continue fine-tuning."
        ),
    )
    parser.add_argument(
        "--upcast_attn",
        default=False,
        type=bool,
        help=(
            "Whether the attention computation should always be upcasted. This is necessary when running stable"
            " diffusion 2.1."
        ),
    )
    parser.add_argument("--dump_path", default=None, type=str, required=True, help="Path to the output model.")
    args = parser.parse_args()

    image_size = args.image_size
    prediction_type = args.prediction_type

    if not os.path.isfile(args.checkpoint_path):
        print(f"Could not find a file at {args.checkpoint_path}")
        exit(-1)

    try:
        checkpoint = load_file(args.checkpoint_path)
    except Exception as e:
        print(e)
        while True:
            print("""
   ######## WARNING: ########
Model did not load as a safetensor. Loading directly via torch.load() may be possible (and required for .ckpts), but carries a security risk!
Unless you are certain that you can trust the source, or have taken necessary precautions and are converting this in an isolated, unprivileged and disposable environment, proceeding is not recommended.
You may want to safely inspect and sanitize the contents of the file first.
"""
            )
            user_in = input("Are you sure? Type 'y' or 'yes' to proceed, or type 'n' or 'no' (or exit the script) to abort conversion. >>> ").lower().strip()
            if user_in in ["n","no"]:
                exit(-1)
            if user_in in ["y","yes"]:
                print("Proceeding.")
                break
        checkpoint = torch.load(args.checkpoint_path)

    # Sometimes models don't have the global_step item
    if "global_step" in checkpoint:
        global_step = checkpoint["global_step"]
    else:
        print("global_step key not found in model")
        global_step = None

    chckpoint_dict_replacements = {
        'cond_stage_model.transformer.embeddings.': 'cond_stage_model.transformer.text_model.embeddings.',
        'cond_stage_model.transformer.encoder.': 'cond_stage_model.transformer.text_model.encoder.',
        'cond_stage_model.transformer.final_layer_norm.': 'cond_stage_model.transformer.text_model.final_layer_norm.',
    }
    def transform_checkpoint_dict_key(k):
        for text, replacement in chckpoint_dict_replacements.items():
            if k.startswith(text):
                k = replacement + k[len(text):]
        return k

    def shared_pointers(tensors):
        from collections import defaultdict
        ptrs = defaultdict(list)
        for k, v in tensors.items():
            ptrs[v.data_ptr()].append(k)
        failing = []
        for ptr, names in ptrs.items():
            if len(names) > 1:
                failing.append(names)
        return failing

    # https://github.com/ratwithacompiler/diffusers_stablediff_conversion/blob/main/convert_diffusers_to_sd.py "original SD key" -> "Diffusers key" Keymap for repairs. "yes this is horrid".
    KeyMap = { "model.diffusion_model.time_embed.0.weight": "time_embedding.linear_1.weight", "model.diffusion_model.time_embed.0.bias": "time_embedding.linear_1.bias", "model.diffusion_model.time_embed.2.weight": "time_embedding.linear_2.weight", "model.diffusion_model.time_embed.2.bias": "time_embedding.linear_2.bias", "model.diffusion_model.input_blocks.0.0.weight": "conv_in.weight", "model.diffusion_model.input_blocks.0.0.bias": "conv_in.bias", "model.diffusion_model.out.0.weight": "conv_norm_out.weight", "model.diffusion_model.out.0.bias": "conv_norm_out.bias", "model.diffusion_model.out.2.weight": "conv_out.weight", "model.diffusion_model.out.2.bias": "conv_out.bias", "model.diffusion_model.input_blocks.1.0.in_layers.0.weight": "down_blocks.0.resnets.0.norm1.weight", "model.diffusion_model.input_blocks.1.0.in_layers.0.bias": "down_blocks.0.resnets.0.norm1.bias", "model.diffusion_model.input_blocks.1.0.in_layers.2.weight": "down_blocks.0.resnets.0.conv1.weight", "model.diffusion_model.input_blocks.1.0.in_layers.2.bias": "down_blocks.0.resnets.0.conv1.bias", "model.diffusion_model.input_blocks.1.0.emb_layers.1.weight": "down_blocks.0.resnets.0.time_emb_proj.weight", "model.diffusion_model.input_blocks.1.0.emb_layers.1.bias": "down_blocks.0.resnets.0.time_emb_proj.bias", "model.diffusion_model.input_blocks.1.0.out_layers.0.weight": "down_blocks.0.resnets.0.norm2.weight", "model.diffusion_model.input_blocks.1.0.out_layers.0.bias": "down_blocks.0.resnets.0.norm2.bias", "model.diffusion_model.input_blocks.1.0.out_layers.3.weight": "down_blocks.0.resnets.0.conv2.weight", "model.diffusion_model.input_blocks.1.0.out_layers.3.bias": "down_blocks.0.resnets.0.conv2.bias", "model.diffusion_model.input_blocks.1.1.norm.weight": "down_blocks.0.attentions.0.norm.weight", "model.diffusion_model.input_blocks.1.1.norm.bias": "down_blocks.0.attentions.0.norm.bias", "model.diffusion_model.input_blocks.1.1.proj_in.weight": "down_blocks.0.attentions.0.proj_in.weight", "model.diffusion_model.input_blocks.1.1.proj_in.bias": "down_blocks.0.attentions.0.proj_in.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.0.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.0.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.0.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm1.weight": "down_blocks.0.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm1.bias": "down_blocks.0.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.weight": "down_blocks.0.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm2.bias": "down_blocks.0.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.weight": "down_blocks.0.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.1.1.transformer_blocks.0.norm3.bias": "down_blocks.0.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.1.1.proj_out.weight": "down_blocks.0.attentions.0.proj_out.weight", "model.diffusion_model.input_blocks.1.1.proj_out.bias": "down_blocks.0.attentions.0.proj_out.bias", "model.diffusion_model.input_blocks.2.0.in_layers.0.weight": "down_blocks.0.resnets.1.norm1.weight", "model.diffusion_model.input_blocks.2.0.in_layers.0.bias": "down_blocks.0.resnets.1.norm1.bias", "model.diffusion_model.input_blocks.2.0.in_layers.2.weight": "down_blocks.0.resnets.1.conv1.weight", "model.diffusion_model.input_blocks.2.0.in_layers.2.bias": "down_blocks.0.resnets.1.conv1.bias", "model.diffusion_model.input_blocks.2.0.emb_layers.1.weight": "down_blocks.0.resnets.1.time_emb_proj.weight", "model.diffusion_model.input_blocks.2.0.emb_layers.1.bias": "down_blocks.0.resnets.1.time_emb_proj.bias", "model.diffusion_model.input_blocks.2.0.out_layers.0.weight": "down_blocks.0.resnets.1.norm2.weight", "model.diffusion_model.input_blocks.2.0.out_layers.0.bias": "down_blocks.0.resnets.1.norm2.bias", "model.diffusion_model.input_blocks.2.0.out_layers.3.weight": "down_blocks.0.resnets.1.conv2.weight", "model.diffusion_model.input_blocks.2.0.out_layers.3.bias": "down_blocks.0.resnets.1.conv2.bias", "model.diffusion_model.input_blocks.2.1.norm.weight": "down_blocks.0.attentions.1.norm.weight", "model.diffusion_model.input_blocks.2.1.norm.bias": "down_blocks.0.attentions.1.norm.bias", "model.diffusion_model.input_blocks.2.1.proj_in.weight": "down_blocks.0.attentions.1.proj_in.weight", "model.diffusion_model.input_blocks.2.1.proj_in.bias": "down_blocks.0.attentions.1.proj_in.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.0.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.0.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.0.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm1.weight": "down_blocks.0.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm1.bias": "down_blocks.0.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm2.weight": "down_blocks.0.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm2.bias": "down_blocks.0.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.weight": "down_blocks.0.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.norm3.bias": "down_blocks.0.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.2.1.proj_out.weight": "down_blocks.0.attentions.1.proj_out.weight", "model.diffusion_model.input_blocks.2.1.proj_out.bias": "down_blocks.0.attentions.1.proj_out.bias", "model.diffusion_model.input_blocks.3.0.op.weight": "down_blocks.0.downsamplers.0.conv.weight", "model.diffusion_model.input_blocks.3.0.op.bias": "down_blocks.0.downsamplers.0.conv.bias", "model.diffusion_model.input_blocks.4.0.in_layers.0.weight": "down_blocks.1.resnets.0.norm1.weight", "model.diffusion_model.input_blocks.4.0.in_layers.0.bias": "down_blocks.1.resnets.0.norm1.bias", "model.diffusion_model.input_blocks.4.0.in_layers.2.weight": "down_blocks.1.resnets.0.conv1.weight", "model.diffusion_model.input_blocks.4.0.in_layers.2.bias": "down_blocks.1.resnets.0.conv1.bias", "model.diffusion_model.input_blocks.4.0.emb_layers.1.weight": "down_blocks.1.resnets.0.time_emb_proj.weight", "model.diffusion_model.input_blocks.4.0.emb_layers.1.bias": "down_blocks.1.resnets.0.time_emb_proj.bias", "model.diffusion_model.input_blocks.4.0.out_layers.0.weight": "down_blocks.1.resnets.0.norm2.weight", "model.diffusion_model.input_blocks.4.0.out_layers.0.bias": "down_blocks.1.resnets.0.norm2.bias", "model.diffusion_model.input_blocks.4.0.out_layers.3.weight": "down_blocks.1.resnets.0.conv2.weight", "model.diffusion_model.input_blocks.4.0.out_layers.3.bias": "down_blocks.1.resnets.0.conv2.bias", "model.diffusion_model.input_blocks.4.0.skip_connection.weight": "down_blocks.1.resnets.0.conv_shortcut.weight", "model.diffusion_model.input_blocks.4.0.skip_connection.bias": "down_blocks.1.resnets.0.conv_shortcut.bias", "model.diffusion_model.input_blocks.4.1.norm.weight": "down_blocks.1.attentions.0.norm.weight", "model.diffusion_model.input_blocks.4.1.norm.bias": "down_blocks.1.attentions.0.norm.bias", "model.diffusion_model.input_blocks.4.1.proj_in.weight": "down_blocks.1.attentions.0.proj_in.weight", "model.diffusion_model.input_blocks.4.1.proj_in.bias": "down_blocks.1.attentions.0.proj_in.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.1.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1.weight": "down_blocks.1.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm1.bias": "down_blocks.1.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.weight": "down_blocks.1.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm2.bias": "down_blocks.1.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.weight": "down_blocks.1.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.4.1.transformer_blocks.0.norm3.bias": "down_blocks.1.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.4.1.proj_out.weight": "down_blocks.1.attentions.0.proj_out.weight", "model.diffusion_model.input_blocks.4.1.proj_out.bias": "down_blocks.1.attentions.0.proj_out.bias", "model.diffusion_model.input_blocks.5.0.in_layers.0.weight": "down_blocks.1.resnets.1.norm1.weight", "model.diffusion_model.input_blocks.5.0.in_layers.0.bias": "down_blocks.1.resnets.1.norm1.bias", "model.diffusion_model.input_blocks.5.0.in_layers.2.weight": "down_blocks.1.resnets.1.conv1.weight", "model.diffusion_model.input_blocks.5.0.in_layers.2.bias": "down_blocks.1.resnets.1.conv1.bias", "model.diffusion_model.input_blocks.5.0.emb_layers.1.weight": "down_blocks.1.resnets.1.time_emb_proj.weight", "model.diffusion_model.input_blocks.5.0.emb_layers.1.bias": "down_blocks.1.resnets.1.time_emb_proj.bias", "model.diffusion_model.input_blocks.5.0.out_layers.0.weight": "down_blocks.1.resnets.1.norm2.weight", "model.diffusion_model.input_blocks.5.0.out_layers.0.bias": "down_blocks.1.resnets.1.norm2.bias", "model.diffusion_model.input_blocks.5.0.out_layers.3.weight": "down_blocks.1.resnets.1.conv2.weight", "model.diffusion_model.input_blocks.5.0.out_layers.3.bias": "down_blocks.1.resnets.1.conv2.bias", "model.diffusion_model.input_blocks.5.1.norm.weight": "down_blocks.1.attentions.1.norm.weight", "model.diffusion_model.input_blocks.5.1.norm.bias": "down_blocks.1.attentions.1.norm.bias", "model.diffusion_model.input_blocks.5.1.proj_in.weight": "down_blocks.1.attentions.1.proj_in.weight", "model.diffusion_model.input_blocks.5.1.proj_in.bias": "down_blocks.1.attentions.1.proj_in.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.1.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1.weight": "down_blocks.1.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm1.bias": "down_blocks.1.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2.weight": "down_blocks.1.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm2.bias": "down_blocks.1.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.weight": "down_blocks.1.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.5.1.transformer_blocks.0.norm3.bias": "down_blocks.1.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.5.1.proj_out.weight": "down_blocks.1.attentions.1.proj_out.weight", "model.diffusion_model.input_blocks.5.1.proj_out.bias": "down_blocks.1.attentions.1.proj_out.bias", "model.diffusion_model.input_blocks.6.0.op.weight": "down_blocks.1.downsamplers.0.conv.weight", "model.diffusion_model.input_blocks.6.0.op.bias": "down_blocks.1.downsamplers.0.conv.bias", "model.diffusion_model.input_blocks.7.0.in_layers.0.weight": "down_blocks.2.resnets.0.norm1.weight", "model.diffusion_model.input_blocks.7.0.in_layers.0.bias": "down_blocks.2.resnets.0.norm1.bias", "model.diffusion_model.input_blocks.7.0.in_layers.2.weight": "down_blocks.2.resnets.0.conv1.weight", "model.diffusion_model.input_blocks.7.0.in_layers.2.bias": "down_blocks.2.resnets.0.conv1.bias", "model.diffusion_model.input_blocks.7.0.emb_layers.1.weight": "down_blocks.2.resnets.0.time_emb_proj.weight", "model.diffusion_model.input_blocks.7.0.emb_layers.1.bias": "down_blocks.2.resnets.0.time_emb_proj.bias", "model.diffusion_model.input_blocks.7.0.out_layers.0.weight": "down_blocks.2.resnets.0.norm2.weight", "model.diffusion_model.input_blocks.7.0.out_layers.0.bias": "down_blocks.2.resnets.0.norm2.bias", "model.diffusion_model.input_blocks.7.0.out_layers.3.weight": "down_blocks.2.resnets.0.conv2.weight", "model.diffusion_model.input_blocks.7.0.out_layers.3.bias": "down_blocks.2.resnets.0.conv2.bias", "model.diffusion_model.input_blocks.7.0.skip_connection.weight": "down_blocks.2.resnets.0.conv_shortcut.weight", "model.diffusion_model.input_blocks.7.0.skip_connection.bias": "down_blocks.2.resnets.0.conv_shortcut.bias", "model.diffusion_model.input_blocks.7.1.norm.weight": "down_blocks.2.attentions.0.norm.weight", "model.diffusion_model.input_blocks.7.1.norm.bias": "down_blocks.2.attentions.0.norm.bias", "model.diffusion_model.input_blocks.7.1.proj_in.weight": "down_blocks.2.attentions.0.proj_in.weight", "model.diffusion_model.input_blocks.7.1.proj_in.bias": "down_blocks.2.attentions.0.proj_in.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.2.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1.weight": "down_blocks.2.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm1.bias": "down_blocks.2.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.weight": "down_blocks.2.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm2.bias": "down_blocks.2.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.weight": "down_blocks.2.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.7.1.transformer_blocks.0.norm3.bias": "down_blocks.2.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.7.1.proj_out.weight": "down_blocks.2.attentions.0.proj_out.weight", "model.diffusion_model.input_blocks.7.1.proj_out.bias": "down_blocks.2.attentions.0.proj_out.bias", "model.diffusion_model.input_blocks.8.0.in_layers.0.weight": "down_blocks.2.resnets.1.norm1.weight", "model.diffusion_model.input_blocks.8.0.in_layers.0.bias": "down_blocks.2.resnets.1.norm1.bias", "model.diffusion_model.input_blocks.8.0.in_layers.2.weight": "down_blocks.2.resnets.1.conv1.weight", "model.diffusion_model.input_blocks.8.0.in_layers.2.bias": "down_blocks.2.resnets.1.conv1.bias", "model.diffusion_model.input_blocks.8.0.emb_layers.1.weight": "down_blocks.2.resnets.1.time_emb_proj.weight", "model.diffusion_model.input_blocks.8.0.emb_layers.1.bias": "down_blocks.2.resnets.1.time_emb_proj.bias", "model.diffusion_model.input_blocks.8.0.out_layers.0.weight": "down_blocks.2.resnets.1.norm2.weight", "model.diffusion_model.input_blocks.8.0.out_layers.0.bias": "down_blocks.2.resnets.1.norm2.bias", "model.diffusion_model.input_blocks.8.0.out_layers.3.weight": "down_blocks.2.resnets.1.conv2.weight", "model.diffusion_model.input_blocks.8.0.out_layers.3.bias": "down_blocks.2.resnets.1.conv2.bias", "model.diffusion_model.input_blocks.8.1.norm.weight": "down_blocks.2.attentions.1.norm.weight", "model.diffusion_model.input_blocks.8.1.norm.bias": "down_blocks.2.attentions.1.norm.bias", "model.diffusion_model.input_blocks.8.1.proj_in.weight": "down_blocks.2.attentions.1.proj_in.weight", "model.diffusion_model.input_blocks.8.1.proj_in.bias": "down_blocks.2.attentions.1.proj_in.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_q.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_k.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_v.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_out.0.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn1.to_out.0.bias": "down_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.0.proj.weight": "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.0.proj.bias": "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.2.weight": "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.ff.net.2.bias": "down_blocks.2.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_q.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_k.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_v.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_out.0.weight": "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.attn2.to_out.0.bias": "down_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm1.weight": "down_blocks.2.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm1.bias": "down_blocks.2.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm2.weight": "down_blocks.2.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm2.bias": "down_blocks.2.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.weight": "down_blocks.2.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.input_blocks.8.1.transformer_blocks.0.norm3.bias": "down_blocks.2.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.input_blocks.8.1.proj_out.weight": "down_blocks.2.attentions.1.proj_out.weight", "model.diffusion_model.input_blocks.8.1.proj_out.bias": "down_blocks.2.attentions.1.proj_out.bias", "model.diffusion_model.input_blocks.9.0.op.weight": "down_blocks.2.downsamplers.0.conv.weight", "model.diffusion_model.input_blocks.9.0.op.bias": "down_blocks.2.downsamplers.0.conv.bias", "model.diffusion_model.input_blocks.10.0.in_layers.0.weight": "down_blocks.3.resnets.0.norm1.weight", "model.diffusion_model.input_blocks.10.0.in_layers.0.bias": "down_blocks.3.resnets.0.norm1.bias", "model.diffusion_model.input_blocks.10.0.in_layers.2.weight": "down_blocks.3.resnets.0.conv1.weight", "model.diffusion_model.input_blocks.10.0.in_layers.2.bias": "down_blocks.3.resnets.0.conv1.bias", "model.diffusion_model.input_blocks.10.0.emb_layers.1.weight": "down_blocks.3.resnets.0.time_emb_proj.weight", "model.diffusion_model.input_blocks.10.0.emb_layers.1.bias": "down_blocks.3.resnets.0.time_emb_proj.bias", "model.diffusion_model.input_blocks.10.0.out_layers.0.weight": "down_blocks.3.resnets.0.norm2.weight", "model.diffusion_model.input_blocks.10.0.out_layers.0.bias": "down_blocks.3.resnets.0.norm2.bias", "model.diffusion_model.input_blocks.10.0.out_layers.3.weight": "down_blocks.3.resnets.0.conv2.weight", "model.diffusion_model.input_blocks.10.0.out_layers.3.bias": "down_blocks.3.resnets.0.conv2.bias", "model.diffusion_model.input_blocks.11.0.in_layers.0.weight": "down_blocks.3.resnets.1.norm1.weight", "model.diffusion_model.input_blocks.11.0.in_layers.0.bias": "down_blocks.3.resnets.1.norm1.bias", "model.diffusion_model.input_blocks.11.0.in_layers.2.weight": "down_blocks.3.resnets.1.conv1.weight", "model.diffusion_model.input_blocks.11.0.in_layers.2.bias": "down_blocks.3.resnets.1.conv1.bias", "model.diffusion_model.input_blocks.11.0.emb_layers.1.weight": "down_blocks.3.resnets.1.time_emb_proj.weight", "model.diffusion_model.input_blocks.11.0.emb_layers.1.bias": "down_blocks.3.resnets.1.time_emb_proj.bias", "model.diffusion_model.input_blocks.11.0.out_layers.0.weight": "down_blocks.3.resnets.1.norm2.weight", "model.diffusion_model.input_blocks.11.0.out_layers.0.bias": "down_blocks.3.resnets.1.norm2.bias", "model.diffusion_model.input_blocks.11.0.out_layers.3.weight": "down_blocks.3.resnets.1.conv2.weight", "model.diffusion_model.input_blocks.11.0.out_layers.3.bias": "down_blocks.3.resnets.1.conv2.bias", "model.diffusion_model.middle_block.0.in_layers.0.weight": "mid_block.resnets.0.norm1.weight", "model.diffusion_model.middle_block.0.in_layers.0.bias": "mid_block.resnets.0.norm1.bias", "model.diffusion_model.middle_block.0.in_layers.2.weight": "mid_block.resnets.0.conv1.weight", "model.diffusion_model.middle_block.0.in_layers.2.bias": "mid_block.resnets.0.conv1.bias", "model.diffusion_model.middle_block.0.emb_layers.1.weight": "mid_block.resnets.0.time_emb_proj.weight", "model.diffusion_model.middle_block.0.emb_layers.1.bias": "mid_block.resnets.0.time_emb_proj.bias", "model.diffusion_model.middle_block.0.out_layers.0.weight": "mid_block.resnets.0.norm2.weight", "model.diffusion_model.middle_block.0.out_layers.0.bias": "mid_block.resnets.0.norm2.bias", "model.diffusion_model.middle_block.0.out_layers.3.weight": "mid_block.resnets.0.conv2.weight", "model.diffusion_model.middle_block.0.out_layers.3.bias": "mid_block.resnets.0.conv2.bias", "model.diffusion_model.middle_block.2.in_layers.0.weight": "mid_block.resnets.1.norm1.weight", "model.diffusion_model.middle_block.2.in_layers.0.bias": "mid_block.resnets.1.norm1.bias", "model.diffusion_model.middle_block.2.in_layers.2.weight": "mid_block.resnets.1.conv1.weight", "model.diffusion_model.middle_block.2.in_layers.2.bias": "mid_block.resnets.1.conv1.bias", "model.diffusion_model.middle_block.2.emb_layers.1.weight": "mid_block.resnets.1.time_emb_proj.weight", "model.diffusion_model.middle_block.2.emb_layers.1.bias": "mid_block.resnets.1.time_emb_proj.bias", "model.diffusion_model.middle_block.2.out_layers.0.weight": "mid_block.resnets.1.norm2.weight", "model.diffusion_model.middle_block.2.out_layers.0.bias": "mid_block.resnets.1.norm2.bias", "model.diffusion_model.middle_block.2.out_layers.3.weight": "mid_block.resnets.1.conv2.weight", "model.diffusion_model.middle_block.2.out_layers.3.bias": "mid_block.resnets.1.conv2.bias", "model.diffusion_model.middle_block.1.norm.weight": "mid_block.attentions.0.norm.weight", "model.diffusion_model.middle_block.1.norm.bias": "mid_block.attentions.0.norm.bias", "model.diffusion_model.middle_block.1.proj_in.weight": "mid_block.attentions.0.proj_in.weight", "model.diffusion_model.middle_block.1.proj_in.bias": "mid_block.attentions.0.proj_in.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_q.weight": "mid_block.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_k.weight": "mid_block.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_v.weight": "mid_block.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_out.0.weight": "mid_block.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn1.to_out.0.bias": "mid_block.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.0.proj.weight": "mid_block.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.0.proj.bias": "mid_block.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.2.weight": "mid_block.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.ff.net.2.bias": "mid_block.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_q.weight": "mid_block.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_k.weight": "mid_block.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_v.weight": "mid_block.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.weight": "mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.attn2.to_out.0.bias": "mid_block.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm1.weight": "mid_block.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm1.bias": "mid_block.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.weight": "mid_block.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm2.bias": "mid_block.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.weight": "mid_block.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.middle_block.1.transformer_blocks.0.norm3.bias": "mid_block.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.middle_block.1.proj_out.weight": "mid_block.attentions.0.proj_out.weight", "model.diffusion_model.middle_block.1.proj_out.bias": "mid_block.attentions.0.proj_out.bias", "model.diffusion_model.output_blocks.0.0.in_layers.0.weight": "up_blocks.0.resnets.0.norm1.weight", "model.diffusion_model.output_blocks.0.0.in_layers.0.bias": "up_blocks.0.resnets.0.norm1.bias", "model.diffusion_model.output_blocks.0.0.in_layers.2.weight": "up_blocks.0.resnets.0.conv1.weight", "model.diffusion_model.output_blocks.0.0.in_layers.2.bias": "up_blocks.0.resnets.0.conv1.bias", "model.diffusion_model.output_blocks.0.0.emb_layers.1.weight": "up_blocks.0.resnets.0.time_emb_proj.weight", "model.diffusion_model.output_blocks.0.0.emb_layers.1.bias": "up_blocks.0.resnets.0.time_emb_proj.bias", "model.diffusion_model.output_blocks.0.0.out_layers.0.weight": "up_blocks.0.resnets.0.norm2.weight", "model.diffusion_model.output_blocks.0.0.out_layers.0.bias": "up_blocks.0.resnets.0.norm2.bias", "model.diffusion_model.output_blocks.0.0.out_layers.3.weight": "up_blocks.0.resnets.0.conv2.weight", "model.diffusion_model.output_blocks.0.0.out_layers.3.bias": "up_blocks.0.resnets.0.conv2.bias", "model.diffusion_model.output_blocks.0.0.skip_connection.weight": "up_blocks.0.resnets.0.conv_shortcut.weight", "model.diffusion_model.output_blocks.0.0.skip_connection.bias": "up_blocks.0.resnets.0.conv_shortcut.bias", "model.diffusion_model.output_blocks.1.0.in_layers.0.weight": "up_blocks.0.resnets.1.norm1.weight", "model.diffusion_model.output_blocks.1.0.in_layers.0.bias": "up_blocks.0.resnets.1.norm1.bias", "model.diffusion_model.output_blocks.1.0.in_layers.2.weight": "up_blocks.0.resnets.1.conv1.weight", "model.diffusion_model.output_blocks.1.0.in_layers.2.bias": "up_blocks.0.resnets.1.conv1.bias", "model.diffusion_model.output_blocks.1.0.emb_layers.1.weight": "up_blocks.0.resnets.1.time_emb_proj.weight", "model.diffusion_model.output_blocks.1.0.emb_layers.1.bias": "up_blocks.0.resnets.1.time_emb_proj.bias", "model.diffusion_model.output_blocks.1.0.out_layers.0.weight": "up_blocks.0.resnets.1.norm2.weight", "model.diffusion_model.output_blocks.1.0.out_layers.0.bias": "up_blocks.0.resnets.1.norm2.bias", "model.diffusion_model.output_blocks.1.0.out_layers.3.weight": "up_blocks.0.resnets.1.conv2.weight", "model.diffusion_model.output_blocks.1.0.out_layers.3.bias": "up_blocks.0.resnets.1.conv2.bias", "model.diffusion_model.output_blocks.1.0.skip_connection.weight": "up_blocks.0.resnets.1.conv_shortcut.weight", "model.diffusion_model.output_blocks.1.0.skip_connection.bias": "up_blocks.0.resnets.1.conv_shortcut.bias", "model.diffusion_model.output_blocks.2.0.in_layers.0.weight": "up_blocks.0.resnets.2.norm1.weight", "model.diffusion_model.output_blocks.2.0.in_layers.0.bias": "up_blocks.0.resnets.2.norm1.bias", "model.diffusion_model.output_blocks.2.0.in_layers.2.weight": "up_blocks.0.resnets.2.conv1.weight", "model.diffusion_model.output_blocks.2.0.in_layers.2.bias": "up_blocks.0.resnets.2.conv1.bias", "model.diffusion_model.output_blocks.2.0.emb_layers.1.weight": "up_blocks.0.resnets.2.time_emb_proj.weight", "model.diffusion_model.output_blocks.2.0.emb_layers.1.bias": "up_blocks.0.resnets.2.time_emb_proj.bias", "model.diffusion_model.output_blocks.2.0.out_layers.0.weight": "up_blocks.0.resnets.2.norm2.weight", "model.diffusion_model.output_blocks.2.0.out_layers.0.bias": "up_blocks.0.resnets.2.norm2.bias", "model.diffusion_model.output_blocks.2.0.out_layers.3.weight": "up_blocks.0.resnets.2.conv2.weight", "model.diffusion_model.output_blocks.2.0.out_layers.3.bias": "up_blocks.0.resnets.2.conv2.bias", "model.diffusion_model.output_blocks.2.0.skip_connection.weight": "up_blocks.0.resnets.2.conv_shortcut.weight", "model.diffusion_model.output_blocks.2.0.skip_connection.bias": "up_blocks.0.resnets.2.conv_shortcut.bias", "model.diffusion_model.output_blocks.2.1.conv.weight": "up_blocks.0.upsamplers.0.conv.weight", "model.diffusion_model.output_blocks.2.1.conv.bias": "up_blocks.0.upsamplers.0.conv.bias", "model.diffusion_model.output_blocks.3.0.in_layers.0.weight": "up_blocks.1.resnets.0.norm1.weight", "model.diffusion_model.output_blocks.3.0.in_layers.0.bias": "up_blocks.1.resnets.0.norm1.bias", "model.diffusion_model.output_blocks.3.0.in_layers.2.weight": "up_blocks.1.resnets.0.conv1.weight", "model.diffusion_model.output_blocks.3.0.in_layers.2.bias": "up_blocks.1.resnets.0.conv1.bias", "model.diffusion_model.output_blocks.3.0.emb_layers.1.weight": "up_blocks.1.resnets.0.time_emb_proj.weight", "model.diffusion_model.output_blocks.3.0.emb_layers.1.bias": "up_blocks.1.resnets.0.time_emb_proj.bias", "model.diffusion_model.output_blocks.3.0.out_layers.0.weight": "up_blocks.1.resnets.0.norm2.weight", "model.diffusion_model.output_blocks.3.0.out_layers.0.bias": "up_blocks.1.resnets.0.norm2.bias", "model.diffusion_model.output_blocks.3.0.out_layers.3.weight": "up_blocks.1.resnets.0.conv2.weight", "model.diffusion_model.output_blocks.3.0.out_layers.3.bias": "up_blocks.1.resnets.0.conv2.bias", "model.diffusion_model.output_blocks.3.0.skip_connection.weight": "up_blocks.1.resnets.0.conv_shortcut.weight", "model.diffusion_model.output_blocks.3.0.skip_connection.bias": "up_blocks.1.resnets.0.conv_shortcut.bias", "model.diffusion_model.output_blocks.3.1.norm.weight": "up_blocks.1.attentions.0.norm.weight", "model.diffusion_model.output_blocks.3.1.norm.bias": "up_blocks.1.attentions.0.norm.bias", "model.diffusion_model.output_blocks.3.1.proj_in.weight": "up_blocks.1.attentions.0.proj_in.weight", "model.diffusion_model.output_blocks.3.1.proj_in.bias": "up_blocks.1.attentions.0.proj_in.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.1.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.1.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.1.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm1.weight": "up_blocks.1.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm1.bias": "up_blocks.1.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm2.weight": "up_blocks.1.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm2.bias": "up_blocks.1.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.weight": "up_blocks.1.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.3.1.transformer_blocks.0.norm3.bias": "up_blocks.1.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.3.1.proj_out.weight": "up_blocks.1.attentions.0.proj_out.weight", "model.diffusion_model.output_blocks.3.1.proj_out.bias": "up_blocks.1.attentions.0.proj_out.bias", "model.diffusion_model.output_blocks.4.0.in_layers.0.weight": "up_blocks.1.resnets.1.norm1.weight", "model.diffusion_model.output_blocks.4.0.in_layers.0.bias": "up_blocks.1.resnets.1.norm1.bias", "model.diffusion_model.output_blocks.4.0.in_layers.2.weight": "up_blocks.1.resnets.1.conv1.weight", "model.diffusion_model.output_blocks.4.0.in_layers.2.bias": "up_blocks.1.resnets.1.conv1.bias", "model.diffusion_model.output_blocks.4.0.emb_layers.1.weight": "up_blocks.1.resnets.1.time_emb_proj.weight", "model.diffusion_model.output_blocks.4.0.emb_layers.1.bias": "up_blocks.1.resnets.1.time_emb_proj.bias", "model.diffusion_model.output_blocks.4.0.out_layers.0.weight": "up_blocks.1.resnets.1.norm2.weight", "model.diffusion_model.output_blocks.4.0.out_layers.0.bias": "up_blocks.1.resnets.1.norm2.bias", "model.diffusion_model.output_blocks.4.0.out_layers.3.weight": "up_blocks.1.resnets.1.conv2.weight", "model.diffusion_model.output_blocks.4.0.out_layers.3.bias": "up_blocks.1.resnets.1.conv2.bias", "model.diffusion_model.output_blocks.4.0.skip_connection.weight": "up_blocks.1.resnets.1.conv_shortcut.weight", "model.diffusion_model.output_blocks.4.0.skip_connection.bias": "up_blocks.1.resnets.1.conv_shortcut.bias", "model.diffusion_model.output_blocks.4.1.norm.weight": "up_blocks.1.attentions.1.norm.weight", "model.diffusion_model.output_blocks.4.1.norm.bias": "up_blocks.1.attentions.1.norm.bias", "model.diffusion_model.output_blocks.4.1.proj_in.weight": "up_blocks.1.attentions.1.proj_in.weight", "model.diffusion_model.output_blocks.4.1.proj_in.bias": "up_blocks.1.attentions.1.proj_in.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.1.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.1.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.1.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm1.weight": "up_blocks.1.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm1.bias": "up_blocks.1.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm2.weight": "up_blocks.1.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm2.bias": "up_blocks.1.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.weight": "up_blocks.1.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.4.1.transformer_blocks.0.norm3.bias": "up_blocks.1.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.4.1.proj_out.weight": "up_blocks.1.attentions.1.proj_out.weight", "model.diffusion_model.output_blocks.4.1.proj_out.bias": "up_blocks.1.attentions.1.proj_out.bias", "model.diffusion_model.output_blocks.5.0.in_layers.0.weight": "up_blocks.1.resnets.2.norm1.weight", "model.diffusion_model.output_blocks.5.0.in_layers.0.bias": "up_blocks.1.resnets.2.norm1.bias", "model.diffusion_model.output_blocks.5.0.in_layers.2.weight": "up_blocks.1.resnets.2.conv1.weight", "model.diffusion_model.output_blocks.5.0.in_layers.2.bias": "up_blocks.1.resnets.2.conv1.bias", "model.diffusion_model.output_blocks.5.0.emb_layers.1.weight": "up_blocks.1.resnets.2.time_emb_proj.weight", "model.diffusion_model.output_blocks.5.0.emb_layers.1.bias": "up_blocks.1.resnets.2.time_emb_proj.bias", "model.diffusion_model.output_blocks.5.0.out_layers.0.weight": "up_blocks.1.resnets.2.norm2.weight", "model.diffusion_model.output_blocks.5.0.out_layers.0.bias": "up_blocks.1.resnets.2.norm2.bias", "model.diffusion_model.output_blocks.5.0.out_layers.3.weight": "up_blocks.1.resnets.2.conv2.weight", "model.diffusion_model.output_blocks.5.0.out_layers.3.bias": "up_blocks.1.resnets.2.conv2.bias", "model.diffusion_model.output_blocks.5.0.skip_connection.weight": "up_blocks.1.resnets.2.conv_shortcut.weight", "model.diffusion_model.output_blocks.5.0.skip_connection.bias": "up_blocks.1.resnets.2.conv_shortcut.bias", "model.diffusion_model.output_blocks.5.2.conv.weight": "up_blocks.1.upsamplers.0.conv.weight", "model.diffusion_model.output_blocks.5.2.conv.bias": "up_blocks.1.upsamplers.0.conv.bias", "model.diffusion_model.output_blocks.5.1.norm.weight": "up_blocks.1.attentions.2.norm.weight", "model.diffusion_model.output_blocks.5.1.norm.bias": "up_blocks.1.attentions.2.norm.bias", "model.diffusion_model.output_blocks.5.1.proj_in.weight": "up_blocks.1.attentions.2.proj_in.weight", "model.diffusion_model.output_blocks.5.1.proj_in.bias": "up_blocks.1.attentions.2.proj_in.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.1.attentions.2.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.1.attentions.2.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.1.attentions.2.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm1.weight": "up_blocks.1.attentions.2.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm1.bias": "up_blocks.1.attentions.2.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm2.weight": "up_blocks.1.attentions.2.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm2.bias": "up_blocks.1.attentions.2.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm3.weight": "up_blocks.1.attentions.2.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.5.1.transformer_blocks.0.norm3.bias": "up_blocks.1.attentions.2.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.5.1.proj_out.weight": "up_blocks.1.attentions.2.proj_out.weight", "model.diffusion_model.output_blocks.5.1.proj_out.bias": "up_blocks.1.attentions.2.proj_out.bias", "model.diffusion_model.output_blocks.6.0.in_layers.0.weight": "up_blocks.2.resnets.0.norm1.weight", "model.diffusion_model.output_blocks.6.0.in_layers.0.bias": "up_blocks.2.resnets.0.norm1.bias", "model.diffusion_model.output_blocks.6.0.in_layers.2.weight": "up_blocks.2.resnets.0.conv1.weight", "model.diffusion_model.output_blocks.6.0.in_layers.2.bias": "up_blocks.2.resnets.0.conv1.bias", "model.diffusion_model.output_blocks.6.0.emb_layers.1.weight": "up_blocks.2.resnets.0.time_emb_proj.weight", "model.diffusion_model.output_blocks.6.0.emb_layers.1.bias": "up_blocks.2.resnets.0.time_emb_proj.bias", "model.diffusion_model.output_blocks.6.0.out_layers.0.weight": "up_blocks.2.resnets.0.norm2.weight", "model.diffusion_model.output_blocks.6.0.out_layers.0.bias": "up_blocks.2.resnets.0.norm2.bias", "model.diffusion_model.output_blocks.6.0.out_layers.3.weight": "up_blocks.2.resnets.0.conv2.weight", "model.diffusion_model.output_blocks.6.0.out_layers.3.bias": "up_blocks.2.resnets.0.conv2.bias", "model.diffusion_model.output_blocks.6.0.skip_connection.weight": "up_blocks.2.resnets.0.conv_shortcut.weight", "model.diffusion_model.output_blocks.6.0.skip_connection.bias": "up_blocks.2.resnets.0.conv_shortcut.bias", "model.diffusion_model.output_blocks.6.1.norm.weight": "up_blocks.2.attentions.0.norm.weight", "model.diffusion_model.output_blocks.6.1.norm.bias": "up_blocks.2.attentions.0.norm.bias", "model.diffusion_model.output_blocks.6.1.proj_in.weight": "up_blocks.2.attentions.0.proj_in.weight", "model.diffusion_model.output_blocks.6.1.proj_in.bias": "up_blocks.2.attentions.0.proj_in.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.2.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.2.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.2.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm1.weight": "up_blocks.2.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm1.bias": "up_blocks.2.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm2.weight": "up_blocks.2.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm2.bias": "up_blocks.2.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.weight": "up_blocks.2.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.6.1.transformer_blocks.0.norm3.bias": "up_blocks.2.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.6.1.proj_out.weight": "up_blocks.2.attentions.0.proj_out.weight", "model.diffusion_model.output_blocks.6.1.proj_out.bias": "up_blocks.2.attentions.0.proj_out.bias", "model.diffusion_model.output_blocks.7.0.in_layers.0.weight": "up_blocks.2.resnets.1.norm1.weight", "model.diffusion_model.output_blocks.7.0.in_layers.0.bias": "up_blocks.2.resnets.1.norm1.bias", "model.diffusion_model.output_blocks.7.0.in_layers.2.weight": "up_blocks.2.resnets.1.conv1.weight", "model.diffusion_model.output_blocks.7.0.in_layers.2.bias": "up_blocks.2.resnets.1.conv1.bias", "model.diffusion_model.output_blocks.7.0.emb_layers.1.weight": "up_blocks.2.resnets.1.time_emb_proj.weight", "model.diffusion_model.output_blocks.7.0.emb_layers.1.bias": "up_blocks.2.resnets.1.time_emb_proj.bias", "model.diffusion_model.output_blocks.7.0.out_layers.0.weight": "up_blocks.2.resnets.1.norm2.weight", "model.diffusion_model.output_blocks.7.0.out_layers.0.bias": "up_blocks.2.resnets.1.norm2.bias", "model.diffusion_model.output_blocks.7.0.out_layers.3.weight": "up_blocks.2.resnets.1.conv2.weight", "model.diffusion_model.output_blocks.7.0.out_layers.3.bias": "up_blocks.2.resnets.1.conv2.bias", "model.diffusion_model.output_blocks.7.0.skip_connection.weight": "up_blocks.2.resnets.1.conv_shortcut.weight", "model.diffusion_model.output_blocks.7.0.skip_connection.bias": "up_blocks.2.resnets.1.conv_shortcut.bias", "model.diffusion_model.output_blocks.7.1.norm.weight": "up_blocks.2.attentions.1.norm.weight", "model.diffusion_model.output_blocks.7.1.norm.bias": "up_blocks.2.attentions.1.norm.bias", "model.diffusion_model.output_blocks.7.1.proj_in.weight": "up_blocks.2.attentions.1.proj_in.weight", "model.diffusion_model.output_blocks.7.1.proj_in.bias": "up_blocks.2.attentions.1.proj_in.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.2.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.2.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.2.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm1.weight": "up_blocks.2.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm1.bias": "up_blocks.2.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm2.weight": "up_blocks.2.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm2.bias": "up_blocks.2.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.weight": "up_blocks.2.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.7.1.transformer_blocks.0.norm3.bias": "up_blocks.2.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.7.1.proj_out.weight": "up_blocks.2.attentions.1.proj_out.weight", "model.diffusion_model.output_blocks.7.1.proj_out.bias": "up_blocks.2.attentions.1.proj_out.bias", "model.diffusion_model.output_blocks.8.0.in_layers.0.weight": "up_blocks.2.resnets.2.norm1.weight", "model.diffusion_model.output_blocks.8.0.in_layers.0.bias": "up_blocks.2.resnets.2.norm1.bias", "model.diffusion_model.output_blocks.8.0.in_layers.2.weight": "up_blocks.2.resnets.2.conv1.weight", "model.diffusion_model.output_blocks.8.0.in_layers.2.bias": "up_blocks.2.resnets.2.conv1.bias", "model.diffusion_model.output_blocks.8.0.emb_layers.1.weight": "up_blocks.2.resnets.2.time_emb_proj.weight", "model.diffusion_model.output_blocks.8.0.emb_layers.1.bias": "up_blocks.2.resnets.2.time_emb_proj.bias", "model.diffusion_model.output_blocks.8.0.out_layers.0.weight": "up_blocks.2.resnets.2.norm2.weight", "model.diffusion_model.output_blocks.8.0.out_layers.0.bias": "up_blocks.2.resnets.2.norm2.bias", "model.diffusion_model.output_blocks.8.0.out_layers.3.weight": "up_blocks.2.resnets.2.conv2.weight", "model.diffusion_model.output_blocks.8.0.out_layers.3.bias": "up_blocks.2.resnets.2.conv2.bias", "model.diffusion_model.output_blocks.8.0.skip_connection.weight": "up_blocks.2.resnets.2.conv_shortcut.weight", "model.diffusion_model.output_blocks.8.0.skip_connection.bias": "up_blocks.2.resnets.2.conv_shortcut.bias", "model.diffusion_model.output_blocks.8.2.conv.weight": "up_blocks.2.upsamplers.0.conv.weight", "model.diffusion_model.output_blocks.8.2.conv.bias": "up_blocks.2.upsamplers.0.conv.bias", "model.diffusion_model.output_blocks.8.1.norm.weight": "up_blocks.2.attentions.2.norm.weight", "model.diffusion_model.output_blocks.8.1.norm.bias": "up_blocks.2.attentions.2.norm.bias", "model.diffusion_model.output_blocks.8.1.proj_in.weight": "up_blocks.2.attentions.2.proj_in.weight", "model.diffusion_model.output_blocks.8.1.proj_in.bias": "up_blocks.2.attentions.2.proj_in.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.2.attentions.2.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.2.attentions.2.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.2.attentions.2.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm1.weight": "up_blocks.2.attentions.2.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm1.bias": "up_blocks.2.attentions.2.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm2.weight": "up_blocks.2.attentions.2.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm2.bias": "up_blocks.2.attentions.2.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm3.weight": "up_blocks.2.attentions.2.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.8.1.transformer_blocks.0.norm3.bias": "up_blocks.2.attentions.2.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.8.1.proj_out.weight": "up_blocks.2.attentions.2.proj_out.weight", "model.diffusion_model.output_blocks.8.1.proj_out.bias": "up_blocks.2.attentions.2.proj_out.bias", "model.diffusion_model.output_blocks.9.0.in_layers.0.weight": "up_blocks.3.resnets.0.norm1.weight", "model.diffusion_model.output_blocks.9.0.in_layers.0.bias": "up_blocks.3.resnets.0.norm1.bias", "model.diffusion_model.output_blocks.9.0.in_layers.2.weight": "up_blocks.3.resnets.0.conv1.weight", "model.diffusion_model.output_blocks.9.0.in_layers.2.bias": "up_blocks.3.resnets.0.conv1.bias", "model.diffusion_model.output_blocks.9.0.emb_layers.1.weight": "up_blocks.3.resnets.0.time_emb_proj.weight", "model.diffusion_model.output_blocks.9.0.emb_layers.1.bias": "up_blocks.3.resnets.0.time_emb_proj.bias", "model.diffusion_model.output_blocks.9.0.out_layers.0.weight": "up_blocks.3.resnets.0.norm2.weight", "model.diffusion_model.output_blocks.9.0.out_layers.0.bias": "up_blocks.3.resnets.0.norm2.bias", "model.diffusion_model.output_blocks.9.0.out_layers.3.weight": "up_blocks.3.resnets.0.conv2.weight", "model.diffusion_model.output_blocks.9.0.out_layers.3.bias": "up_blocks.3.resnets.0.conv2.bias", "model.diffusion_model.output_blocks.9.0.skip_connection.weight": "up_blocks.3.resnets.0.conv_shortcut.weight", "model.diffusion_model.output_blocks.9.0.skip_connection.bias": "up_blocks.3.resnets.0.conv_shortcut.bias", "model.diffusion_model.output_blocks.9.1.norm.weight": "up_blocks.3.attentions.0.norm.weight", "model.diffusion_model.output_blocks.9.1.norm.bias": "up_blocks.3.attentions.0.norm.bias", "model.diffusion_model.output_blocks.9.1.proj_in.weight": "up_blocks.3.attentions.0.proj_in.weight", "model.diffusion_model.output_blocks.9.1.proj_in.bias": "up_blocks.3.attentions.0.proj_in.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.3.attentions.0.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.3.attentions.0.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.3.attentions.0.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm1.weight": "up_blocks.3.attentions.0.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm1.bias": "up_blocks.3.attentions.0.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm2.weight": "up_blocks.3.attentions.0.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm2.bias": "up_blocks.3.attentions.0.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm3.weight": "up_blocks.3.attentions.0.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.9.1.transformer_blocks.0.norm3.bias": "up_blocks.3.attentions.0.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.9.1.proj_out.weight": "up_blocks.3.attentions.0.proj_out.weight", "model.diffusion_model.output_blocks.9.1.proj_out.bias": "up_blocks.3.attentions.0.proj_out.bias", "model.diffusion_model.output_blocks.10.0.in_layers.0.weight": "up_blocks.3.resnets.1.norm1.weight", "model.diffusion_model.output_blocks.10.0.in_layers.0.bias": "up_blocks.3.resnets.1.norm1.bias", "model.diffusion_model.output_blocks.10.0.in_layers.2.weight": "up_blocks.3.resnets.1.conv1.weight", "model.diffusion_model.output_blocks.10.0.in_layers.2.bias": "up_blocks.3.resnets.1.conv1.bias", "model.diffusion_model.output_blocks.10.0.emb_layers.1.weight": "up_blocks.3.resnets.1.time_emb_proj.weight", "model.diffusion_model.output_blocks.10.0.emb_layers.1.bias": "up_blocks.3.resnets.1.time_emb_proj.bias", "model.diffusion_model.output_blocks.10.0.out_layers.0.weight": "up_blocks.3.resnets.1.norm2.weight", "model.diffusion_model.output_blocks.10.0.out_layers.0.bias": "up_blocks.3.resnets.1.norm2.bias", "model.diffusion_model.output_blocks.10.0.out_layers.3.weight": "up_blocks.3.resnets.1.conv2.weight", "model.diffusion_model.output_blocks.10.0.out_layers.3.bias": "up_blocks.3.resnets.1.conv2.bias", "model.diffusion_model.output_blocks.10.0.skip_connection.weight": "up_blocks.3.resnets.1.conv_shortcut.weight", "model.diffusion_model.output_blocks.10.0.skip_connection.bias": "up_blocks.3.resnets.1.conv_shortcut.bias", "model.diffusion_model.output_blocks.10.1.norm.weight": "up_blocks.3.attentions.1.norm.weight", "model.diffusion_model.output_blocks.10.1.norm.bias": "up_blocks.3.attentions.1.norm.bias", "model.diffusion_model.output_blocks.10.1.proj_in.weight": "up_blocks.3.attentions.1.proj_in.weight", "model.diffusion_model.output_blocks.10.1.proj_in.bias": "up_blocks.3.attentions.1.proj_in.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.3.attentions.1.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.3.attentions.1.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.3.attentions.1.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm1.weight": "up_blocks.3.attentions.1.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm1.bias": "up_blocks.3.attentions.1.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm2.weight": "up_blocks.3.attentions.1.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm2.bias": "up_blocks.3.attentions.1.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.weight": "up_blocks.3.attentions.1.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.10.1.transformer_blocks.0.norm3.bias": "up_blocks.3.attentions.1.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.10.1.proj_out.weight": "up_blocks.3.attentions.1.proj_out.weight", "model.diffusion_model.output_blocks.10.1.proj_out.bias": "up_blocks.3.attentions.1.proj_out.bias", "model.diffusion_model.output_blocks.11.0.in_layers.0.weight": "up_blocks.3.resnets.2.norm1.weight", "model.diffusion_model.output_blocks.11.0.in_layers.0.bias": "up_blocks.3.resnets.2.norm1.bias", "model.diffusion_model.output_blocks.11.0.in_layers.2.weight": "up_blocks.3.resnets.2.conv1.weight", "model.diffusion_model.output_blocks.11.0.in_layers.2.bias": "up_blocks.3.resnets.2.conv1.bias", "model.diffusion_model.output_blocks.11.0.emb_layers.1.weight": "up_blocks.3.resnets.2.time_emb_proj.weight", "model.diffusion_model.output_blocks.11.0.emb_layers.1.bias": "up_blocks.3.resnets.2.time_emb_proj.bias", "model.diffusion_model.output_blocks.11.0.out_layers.0.weight": "up_blocks.3.resnets.2.norm2.weight", "model.diffusion_model.output_blocks.11.0.out_layers.0.bias": "up_blocks.3.resnets.2.norm2.bias", "model.diffusion_model.output_blocks.11.0.out_layers.3.weight": "up_blocks.3.resnets.2.conv2.weight", "model.diffusion_model.output_blocks.11.0.out_layers.3.bias": "up_blocks.3.resnets.2.conv2.bias", "model.diffusion_model.output_blocks.11.0.skip_connection.weight": "up_blocks.3.resnets.2.conv_shortcut.weight", "model.diffusion_model.output_blocks.11.0.skip_connection.bias": "up_blocks.3.resnets.2.conv_shortcut.bias", "model.diffusion_model.output_blocks.11.1.norm.weight": "up_blocks.3.attentions.2.norm.weight", "model.diffusion_model.output_blocks.11.1.norm.bias": "up_blocks.3.attentions.2.norm.bias", "model.diffusion_model.output_blocks.11.1.proj_in.weight": "up_blocks.3.attentions.2.proj_in.weight", "model.diffusion_model.output_blocks.11.1.proj_in.bias": "up_blocks.3.attentions.2.proj_in.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn1.to_q.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_q.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn1.to_k.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_k.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn1.to_v.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_v.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn1.to_out.0.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_out.0.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn1.to_out.0.bias": "up_blocks.3.attentions.2.transformer_blocks.0.attn1.to_out.0.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.ff.net.0.proj.weight": "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.0.proj.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.ff.net.0.proj.bias": "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.0.proj.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.ff.net.2.weight": "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.2.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.ff.net.2.bias": "up_blocks.3.attentions.2.transformer_blocks.0.ff.net.2.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_q.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_q.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_k.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_k.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_v.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_v.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_out.0.weight": "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_out.0.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.attn2.to_out.0.bias": "up_blocks.3.attentions.2.transformer_blocks.0.attn2.to_out.0.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm1.weight": "up_blocks.3.attentions.2.transformer_blocks.0.norm1.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm1.bias": "up_blocks.3.attentions.2.transformer_blocks.0.norm1.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm2.weight": "up_blocks.3.attentions.2.transformer_blocks.0.norm2.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm2.bias": "up_blocks.3.attentions.2.transformer_blocks.0.norm2.bias", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.weight": "up_blocks.3.attentions.2.transformer_blocks.0.norm3.weight", "model.diffusion_model.output_blocks.11.1.transformer_blocks.0.norm3.bias": "up_blocks.3.attentions.2.transformer_blocks.0.norm3.bias", "model.diffusion_model.output_blocks.11.1.proj_out.weight": "up_blocks.3.attentions.2.proj_out.weight", "model.diffusion_model.output_blocks.11.1.proj_out.bias": "up_blocks.3.attentions.2.proj_out.bias"}

    # unpack state_dict if encapsulated.
    checkpoint = checkpoint.pop("state_dict", checkpoint)

    # keys known to be missing in some cases
    missing_diffusers_keys_unet = ['up_blocks.0.upsamplers.0.conv.weight', 'up_blocks.0.upsamplers.0.conv.bias', 'up_blocks.1.upsamplers.0.conv.weight', 'up_blocks.1.upsamplers.0.conv.bias', 'up_blocks.2.upsamplers.0.conv.weight', 'up_blocks.2.upsamplers.0.conv.bias']
    # keys known to be incorrectly loaded in some cases
    unwanted_diffusers_keys_unet = ["up_blocks.0.attentions.2.conv.bias", "up_blocks.0.attentions.2.conv.weight"]
    # resolve keys from KeyMap
    missing_keypairs = [([oldkey for oldkey,newkey in KeyMap.items() if newkey==missing_key][0], missing_key) for missing_key in missing_diffusers_keys_unet]
    # store missing keys!
    missing_keys_store = {pair[1]:checkpoint[pair[0]].clone() for pair in missing_keypairs}

    sd = {}
    for k, v in checkpoint.items():
        new_key = transform_checkpoint_dict_key(k)
        if new_key is not None:
            sd[new_key] = v
    checkpoint.clear()
    checkpoint.update(sd)

    print(f"Shared pointers in checkpoint: {shared_pointers(checkpoint)}")

    upcast_attention = False
    if args.original_config_file is None:
        key_name = "model.diffusion_model.input_blocks.2.1.transformer_blocks.0.attn2.to_k.weight"

        if key_name in checkpoint and checkpoint[key_name].shape[-1] == 1024:
            if not os.path.isfile("v2-inference-v.yaml"):
                # model_type = "v2"
                try:
                    import urllib.request
                    urllib.request.urlretrieve("https://raw.githubusercontent.com/Stability-AI/stablediffusion/main/configs/stable-diffusion/v2-inference-v.yaml", "v2-inference-v.yaml")
                except ImportError:
                    print("could not import urllib. attempting wget.")
                    os.system(
                        "wget https://raw.githubusercontent.com/Stability-AI/stablediffusion/main/configs/stable-diffusion/v2-inference-v.yaml"
                        " -O v2-inference-v.yaml"
                    )
            args.original_config_file = "./v2-inference-v.yaml"

            if global_step == 110000:
                # v2.1 needs to upcast attention
                upcast_attention = True
        else:
            if not os.path.isfile("v1-inference.yaml"):
                # model_type = "v1"
                try:
                    import urllib.request
                    urllib.request.urlretrieve("https://raw.githubusercontent.com/CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml", "v1-inference.yaml")
                except ImportError:
                    print("could not import urllib. attempting wget.")
                    os.system(
                        "wget https://raw.githubusercontent.com/CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml"
                        " -O v1-inference.yaml"
                    )
            args.original_config_file = "./v1-inference.yaml"

    original_config = OmegaConf.load(args.original_config_file)

    if args.num_in_channels is not None:
        original_config["model"]["params"]["unet_config"]["params"]["in_channels"] = args.num_in_channels

    if (
        "parameterization" in original_config["model"]["params"]
        and original_config["model"]["params"]["parameterization"] == "v"
    ):
        if prediction_type is None:
            # NOTE: For stable diffusion 2 base it is recommended to pass `prediction_type=="epsilon"`
            # as it relies on a brittle global step parameter here
            prediction_type = "epsilon" if global_step == 875000 else "v_prediction"
        if image_size is None:
            # NOTE: For stable diffusion 2 base one has to pass `image_size==512`
            # as it relies on a brittle global step parameter here
            image_size = 512 if global_step == 875000 else 768
    else:
        if prediction_type is None:
            prediction_type = "epsilon"
        if image_size is None:
            image_size = 512

    num_train_timesteps = original_config.model.params.timesteps
    beta_start = original_config.model.params.linear_start
    beta_end = original_config.model.params.linear_end

    scheduler = DDIMScheduler(
        beta_end=beta_end,
        beta_schedule="scaled_linear",
        beta_start=beta_start,
        num_train_timesteps=num_train_timesteps,
        steps_offset=1,
        clip_sample=False,
        set_alpha_to_one=False,
        prediction_type=prediction_type,
    )
    # make sure scheduler works correctly with DDIM
    scheduler.register_to_config(clip_sample=False)

    if args.scheduler_type == "pndm":
        config = dict(scheduler.config)
        config["skip_prk_steps"] = True
        scheduler = PNDMScheduler.from_config(config)
    elif args.scheduler_type == "lms":
        scheduler = LMSDiscreteScheduler.from_config(scheduler.config)
    #elif args.scheduler_type == "heun":
    #    scheduler = HeunDiscreteScheduler.from_config(scheduler.config)
    elif args.scheduler_type == "euler":
        scheduler = EulerDiscreteScheduler.from_config(scheduler.config)
    #elif args.scheduler_type == "euler-ancestral":
    #    scheduler = EulerAncestralDiscreteScheduler.from_config(scheduler.config)
    #elif args.scheduler_type == "dpm":
    #    scheduler = DPMSolverMultistepScheduler.from_config(scheduler.config)
    elif args.scheduler_type == "ddim":
        scheduler = scheduler
    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, image_size=image_size)
    """unet_config = { # diffusers SD1.4 UNET OVERRIDE
        "_class_name": "UNet2DConditionModel",  "_diffusers_version": "0.6.0",  "act_fn": "silu",  "attention_head_dim": 8,  "block_out_channels": [    320,    640,    1280,    1280  ],
        "center_input_sample": False,  "cross_attention_dim": 768,  "down_block_types": [    "CrossAttnDownBlock2D",    "CrossAttnDownBlock2D",    "CrossAttnDownBlock2D",    "DownBlock2D"  ],
        "downsample_padding": 1,  "flip_sin_to_cos": True,  "freq_shift": 0,  "in_channels": 4,  "layers_per_block": 2,  "mid_block_scale_factor": 1,  "norm_eps": 1e-05,  "norm_num_groups": 32,  "out_channels": 4,  "sample_size": 64,
        "up_block_types": [    "UpBlock2D",    "CrossAttnUpBlock2D",    "CrossAttnUpBlock2D",    "CrossAttnUpBlock2D"  ]
    }"""

    unet_config["upcast_attention"] = upcast_attention
    unet = UNet2DConditionModel(**unet_config)

    converted_unet_checkpoint = convert_ldm_unet_checkpoint(
        checkpoint, unet_config, path=args.checkpoint_path, extract_ema=args.extract_ema
    )

    amended = 0
    for k in missing_keys_store:
        if not k in converted_unet_checkpoint:
            converted_unet_checkpoint[k] = missing_keys_store[k]
            amended += 1
    removed = 0
    for k in unwanted_diffusers_keys_unet:
        if k in converted_unet_checkpoint:
            converted_unet_checkpoint.pop(k)
            removed += 1

    print(f"Of {len(missing_keys_store)} keys known to sometimes be missing for (thus far) unknown reasons, {amended} were missing and amended.")
    print(f"Of {len(unwanted_diffusers_keys_unet)} unsuitable keys known to sometimes be present for (thus far) unknown reasons, {removed} were found and purged.")

    unet.load_state_dict(converted_unet_checkpoint)

    # Convert the VAE model.
    vae_config = create_vae_diffusers_config(original_config, image_size=image_size)
    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.
    model_type = args.pipeline_type
    if model_type is None:
        model_type = original_config.model.params.cond_stage_config.target.split(".")[-1]

    if model_type == "FrozenOpenCLIPEmbedder":
        text_model = convert_open_clip_checkpoint(checkpoint)
        tokenizer = CLIPTokenizer.from_pretrained("stabilityai/stable-diffusion-2", subfolder="tokenizer")
        pipe = StableDiffusionPipeline(
            vae=vae,
            text_encoder=text_model,
            tokenizer=tokenizer,
            unet=unet,
            scheduler=scheduler,
            safety_checker=None,
            feature_extractor=None,
            requires_safety_checker=False,
        )
    elif model_type == "PaintByExample":
        vision_model = convert_paint_by_example_checkpoint(checkpoint)
        tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-large-patch14")
        feature_extractor = AutoFeatureExtractor.from_pretrained("CompVis/stable-diffusion-safety-checker")
        pipe = PaintByExamplePipeline(
            vae=vae,
            image_encoder=vision_model,
            unet=unet,
            scheduler=scheduler,
            safety_checker=None,
            feature_extractor=feature_extractor,
        )
    elif model_type == "FrozenCLIPEmbedder":
        text_model = convert_ldm_clip_checkpoint(checkpoint)
        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)