Convert DreamBooth .bin files to a .ckpt file

convertToCheckpoint.md

Converting DreamBooth .bin files to a .ckpt model file.

These instructions are based on DreamBooth usage with the https://github.com/ShivamShrirao/diffusers repo.

1. Add the script files

Below are 2 files. "convertToCkpt.py" and "toCkpt.sh". Create those files inside the examples/dreambooth folder with the code provided.

1a. Python convert script (required)

Put the convertToCkpt.py file in the examples/dreambooth folder. Credit to @jachiam for this Python script https://gist.github.com/jachiam/8a5c0b607e38fcc585168b90c686eb05

1b. Convenience CLI command (optional)

This runs the Python script. It accepts the model folder as the only argument. It's a little easier to type and automatically uses the folder name as the .ckpt filename.

Put the toCkpt.sh file in the examples/dreambooth folder as well.

2. Running the script

2a. Make sure you're in the examples/dreambooth folder

Run if you're still in the project root directory.

cd examples/dreambooth

2b. Either run the Python script directly, or run the convenience CLI script

The convenience CLI script:

./toCkpt.sh ./name_of_model_folder

Remember to change ./name_of_model_folder to the correct folder name.

If you want to run the original Python script, that can still be done:

python convertToCkpt.py --model_path ./name_of_model_folder --checkpoint_path ./model.ckpt

Remember to change ./name_of_model_folder to the correct folder name. model.ckpt can be anything.

3. Try out your new checkpoint model.

If you're using Automatic1111, copy-paste that .ckpt model file into the models/Stable-diffusion folder.

convertToCkpt.py

# Script for converting a HF Diffusers saved pipeline to a Stable Diffusion checkpoint.
# *Only* converts the UNet, VAE, and Text Encoder.
# Does not convert optimizer state or any other thing.
# Written by jachiam

import argparse
import os.path as osp

import torch


# =================#
# UNet Conversion #
# =================#

unet_conversion_map = [
    # (stable-diffusion, HF Diffusers)
    ("time_embed.0.weight", "time_embedding.linear_1.weight"),
    ("time_embed.0.bias", "time_embedding.linear_1.bias"),
    ("time_embed.2.weight", "time_embedding.linear_2.weight"),
    ("time_embed.2.bias", "time_embedding.linear_2.bias"),
    ("input_blocks.0.0.weight", "conv_in.weight"),
    ("input_blocks.0.0.bias", "conv_in.bias"),
    ("out.0.weight", "conv_norm_out.weight"),
    ("out.0.bias", "conv_norm_out.bias"),
    ("out.2.weight", "conv_out.weight"),
    ("out.2.bias", "conv_out.bias"),
]

unet_conversion_map_resnet = [
    # (stable-diffusion, HF Diffusers)
    ("in_layers.0", "norm1"),
    ("in_layers.2", "conv1"),
    ("out_layers.0", "norm2"),
    ("out_layers.3", "conv2"),
    ("emb_layers.1", "time_emb_proj"),
    ("skip_connection", "conv_shortcut"),
]

unet_conversion_map_layer = []
# hardcoded number of downblocks and resnets/attentions...
# would need smarter logic for other networks.
for i in range(4):
    # loop over downblocks/upblocks

    for j in range(2):
        # loop over resnets/attentions for downblocks
        hf_down_res_prefix = f"down_blocks.{i}.resnets.{j}."
        sd_down_res_prefix = f"input_blocks.{3*i + j + 1}.0."
        unet_conversion_map_layer.append((sd_down_res_prefix, hf_down_res_prefix))

        if i < 3:
            # no attention layers in down_blocks.3
            hf_down_atn_prefix = f"down_blocks.{i}.attentions.{j}."
            sd_down_atn_prefix = f"input_blocks.{3*i + j + 1}.1."
            unet_conversion_map_layer.append((sd_down_atn_prefix, hf_down_atn_prefix))

    for j in range(3):
        # loop over resnets/attentions for upblocks
        hf_up_res_prefix = f"up_blocks.{i}.resnets.{j}."
        sd_up_res_prefix = f"output_blocks.{3*i + j}.0."
        unet_conversion_map_layer.append((sd_up_res_prefix, hf_up_res_prefix))

        if i > 0:
            # no attention layers in up_blocks.0
            hf_up_atn_prefix = f"up_blocks.{i}.attentions.{j}."
            sd_up_atn_prefix = f"output_blocks.{3*i + j}.1."
            unet_conversion_map_layer.append((sd_up_atn_prefix, hf_up_atn_prefix))

    if i < 3:
        # no downsample in down_blocks.3
        hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0.conv."
        sd_downsample_prefix = f"input_blocks.{3*(i+1)}.0.op."
        unet_conversion_map_layer.append((sd_downsample_prefix, hf_downsample_prefix))

        # no upsample in up_blocks.3
        hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
        sd_upsample_prefix = f"output_blocks.{3*i + 2}.{1 if i == 0 else 2}."
        unet_conversion_map_layer.append((sd_upsample_prefix, hf_upsample_prefix))

hf_mid_atn_prefix = "mid_block.attentions.0."
sd_mid_atn_prefix = "middle_block.1."
unet_conversion_map_layer.append((sd_mid_atn_prefix, hf_mid_atn_prefix))

for j in range(2):
    hf_mid_res_prefix = f"mid_block.resnets.{j}."
    sd_mid_res_prefix = f"middle_block.{2*j}."
    unet_conversion_map_layer.append((sd_mid_res_prefix, hf_mid_res_prefix))


def convert_unet_state_dict(unet_state_dict):
    # buyer beware: this is a *brittle* function,
    # and correct output requires that all of these pieces interact in
    # the exact order in which I have arranged them.
    mapping = {k: k for k in unet_state_dict.keys()}
    for sd_name, hf_name in unet_conversion_map:
        mapping[hf_name] = sd_name
    for k, v in mapping.items():
        if "resnets" in k:
            for sd_part, hf_part in unet_conversion_map_resnet:
                v = v.replace(hf_part, sd_part)
            mapping[k] = v
    for k, v in mapping.items():
        for sd_part, hf_part in unet_conversion_map_layer:
            v = v.replace(hf_part, sd_part)
        mapping[k] = v
    new_state_dict = {v: unet_state_dict[k] for k, v in mapping.items()}
    return new_state_dict


# ================#
# VAE Conversion #
# ================#

vae_conversion_map = [
    # (stable-diffusion, HF Diffusers)
    ("nin_shortcut", "conv_shortcut"),
    ("norm_out", "conv_norm_out"),
    ("mid.attn_1.", "mid_block.attentions.0."),
]

for i in range(4):
    # down_blocks have two resnets
    for j in range(2):
        hf_down_prefix = f"encoder.down_blocks.{i}.resnets.{j}."
        sd_down_prefix = f"encoder.down.{i}.block.{j}."
        vae_conversion_map.append((sd_down_prefix, hf_down_prefix))

    if i < 3:
        hf_downsample_prefix = f"down_blocks.{i}.downsamplers.0."
        sd_downsample_prefix = f"down.{i}.downsample."
        vae_conversion_map.append((sd_downsample_prefix, hf_downsample_prefix))

        hf_upsample_prefix = f"up_blocks.{i}.upsamplers.0."
        sd_upsample_prefix = f"up.{3-i}.upsample."
        vae_conversion_map.append((sd_upsample_prefix, hf_upsample_prefix))

    # up_blocks have three resnets
    # also, up blocks in hf are numbered in reverse from sd
    for j in range(3):
        hf_up_prefix = f"decoder.up_blocks.{i}.resnets.{j}."
        sd_up_prefix = f"decoder.up.{3-i}.block.{j}."
        vae_conversion_map.append((sd_up_prefix, hf_up_prefix))

# this part accounts for mid blocks in both the encoder and the decoder
for i in range(2):
    hf_mid_res_prefix = f"mid_block.resnets.{i}."
    sd_mid_res_prefix = f"mid.block_{i+1}."
    vae_conversion_map.append((sd_mid_res_prefix, hf_mid_res_prefix))


vae_conversion_map_attn = [
    # (stable-diffusion, HF Diffusers)
    ("norm.", "group_norm."),
    ("q.", "query."),
    ("k.", "key."),
    ("v.", "value."),
    ("proj_out.", "proj_attn."),
]


def reshape_weight_for_sd(w):
    # convert HF linear weights to SD conv2d weights
    return w.reshape(*w.shape, 1, 1)


def convert_vae_state_dict(vae_state_dict):
    mapping = {k: k for k in vae_state_dict.keys()}
    for k, v in mapping.items():
        for sd_part, hf_part in vae_conversion_map:
            v = v.replace(hf_part, sd_part)
        mapping[k] = v
    for k, v in mapping.items():
        if "attentions" in k:
            for sd_part, hf_part in vae_conversion_map_attn:
                v = v.replace(hf_part, sd_part)
            mapping[k] = v
    new_state_dict = {v: vae_state_dict[k] for k, v in mapping.items()}
    weights_to_convert = ["q", "k", "v", "proj_out"]
    for k, v in new_state_dict.items():
        for weight_name in weights_to_convert:
            if f"mid.attn_1.{weight_name}.weight" in k:
                print(f"Reshaping {k} for SD format")
                new_state_dict[k] = reshape_weight_for_sd(v)
    return new_state_dict


# =========================#
# Text Encoder Conversion #
# =========================#
# pretty much a no-op


def convert_text_enc_state_dict(text_enc_dict):
    return text_enc_dict


if __name__ == "__main__":
    parser = argparse.ArgumentParser()

    parser.add_argument("--model_path", default=None, type=str, required=True, help="Path to the model to convert.")
    parser.add_argument("--checkpoint_path", default=None, type=str, required=True, help="Path to the output model.")
    parser.add_argument("--half", action="store_true", help="Save weights in half precision.")

    args = parser.parse_args()

    assert args.model_path is not None, "Must provide a model path!"

    assert args.checkpoint_path is not None, "Must provide a checkpoint path!"

    unet_path = osp.join(args.model_path, "unet", "diffusion_pytorch_model.bin")
    vae_path = osp.join(args.model_path, "vae", "diffusion_pytorch_model.bin")
    text_enc_path = osp.join(args.model_path, "text_encoder", "pytorch_model.bin")

    # Convert the UNet model
    unet_state_dict = torch.load(unet_path, map_location='cpu')
    unet_state_dict = convert_unet_state_dict(unet_state_dict)
    unet_state_dict = {"model.diffusion_model." + k: v for k, v in unet_state_dict.items()}

    # Convert the VAE model
    vae_state_dict = torch.load(vae_path, map_location='cpu')
    vae_state_dict = convert_vae_state_dict(vae_state_dict)
    vae_state_dict = {"first_stage_model." + k: v for k, v in vae_state_dict.items()}

    # Convert the text encoder model
    text_enc_dict = torch.load(text_enc_path, map_location='cpu')
    text_enc_dict = convert_text_enc_state_dict(text_enc_dict)
    text_enc_dict = {"cond_stage_model.transformer." + k: v for k, v in text_enc_dict.items()}

    # Put together new checkpoint
    state_dict = {**unet_state_dict, **vae_state_dict, **text_enc_dict}
    if args.half:
        state_dict = {k:v.half() for k,v in state_dict.items()}
    state_dict = {"state_dict": state_dict}
    torch.save(state_dict, args.checkpoint_path)

toCkpt.sh

#!/bin/bash

model_path=$1
ckpt_name=$(basename $model_path)
ckpt_path="${ckpt_name}.ckpt"

python convertToCkpt.py --model_path=$model_path --checkpoint_path=$ckpt_path

@AmicusX Read the instructions.

These instructions are based on DreamBooth usage with the https://github.com/ShivamShrirao/diffusers repo.
[...]
Put the toCkpt.sh file in the examples/dreambooth folder as well.

That's this folder (https://github.com/ShivamShrirao/diffusers/tree/main/examples/dreambooth) after you've cloned the repo. You're creating a toCkpt.sh file in <path-of-ShivamShrirao-diffusers-repo>/examples/dreambooth. After creating a custom model in DreamBooth, you get a new folder. You just need to run toCkpt.sh <new-folder-name>.

At the moment, some fundamental knowledge of using the terminal is required. A GUI would be better, but everyone is working on StableDiffusion tools in their free time, so if you have time to build a GUI for custom model creation, then go ahead and share it with the community.

@2blackbar if you find it's easier to write out the Python arguments, that's fine. But, no need to go around complaining on other people's gists.

Is this specific to dreambooth? Im running Diffusers via an IDE, and it dont seem to have a way to save it as a single file / checkpoint. It can do a "save_pretrained" wich i asume saves it as a pipeline / diffusers model, is that the same thing?

doesn't seem to work with .safetensors