Upload folder using huggingface_hub

output/checkpoint-10000/controlnet/config.json

@@ -0,0 +1,57 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.32.0.dev0",
+  "act_fn": "silu",
+  "addition_embed_type": "text_time",
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": 256,
+  "attention_head_dim": [
+    5,
+    10,
+    20
+  ],
+  "block_out_channels": [
+    320,
+    640,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 2048,
+  "down_block_types": [
+    "DownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": 2816,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": [
+    1,
+    2,
+    10
+  ],
+  "upcast_attention": null,
+  "use_linear_projection": true
+}

output/checkpoint-10000/controlnet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bafd13f4d2417e5b3822d1a059cbc196d2abd300768b123941e023ebad915333
+size 5004167864

output/checkpoint-10000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd4904554628f145f67896b54aa980489ad58b54767f41d7ef74ea798673c1b5
+size 10008841510

output/checkpoint-10000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25975d039a2387e37c1aecad5978e16cfe0ff0d0cf34ed90051653adb4e44888
+size 14344

output/checkpoint-10000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d49367aa0f19981d0484e3ef5c1b009e6bdd8e5c0d8ef08f2f8d235b9f843816
+size 988

output/checkpoint-10000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d31ad9a5150e341b638f838a47b3cb3db0d48efa493bcc8d1f8bf787b2c8cee3
+size 1000

output/checkpoint-20000/controlnet/config.json

@@ -0,0 +1,57 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.32.0.dev0",
+  "act_fn": "silu",
+  "addition_embed_type": "text_time",
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": 256,
+  "attention_head_dim": [
+    5,
+    10,
+    20
+  ],
+  "block_out_channels": [
+    320,
+    640,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 2048,
+  "down_block_types": [
+    "DownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": 2816,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": [
+    1,
+    2,
+    10
+  ],
+  "upcast_attention": null,
+  "use_linear_projection": true
+}

output/checkpoint-20000/controlnet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2345c59cd1295084f1e8ed2fa331371806cb7450532f6f813a0d294814f43d1
+size 5004167864

output/checkpoint-20000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b234645d7f7a0b52965799e39e7e858aa22f281da49f3545aea48e3da6c9fda
+size 10008841510

output/checkpoint-20000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37e2fbb65cb9637df6850b605e724fa082972239ff87a4f9918779aa2b77e3d6
+size 14344

output/checkpoint-20000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a33b95288c044d4b8931227377713cfbcc8d73a1b4e184632c562d7fdaf703f9
+size 988

output/checkpoint-20000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fb423915b7f7abbe2676b44b9381dc20a834b3ca26d3e2d87b59cdd536701fe
+size 1000

output/checkpoint-30000/controlnet/config.json

@@ -0,0 +1,57 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.32.0.dev0",
+  "act_fn": "silu",
+  "addition_embed_type": "text_time",
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": 256,
+  "attention_head_dim": [
+    5,
+    10,
+    20
+  ],
+  "block_out_channels": [
+    320,
+    640,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 2048,
+  "down_block_types": [
+    "DownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": 2816,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": [
+    1,
+    2,
+    10
+  ],
+  "upcast_attention": null,
+  "use_linear_projection": true
+}

output/checkpoint-30000/controlnet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2f682210750d2b1f9884f885692856d9bf5559834a9c18f906cb17d956e8ab0
+size 5004167864

output/checkpoint-30000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99e653650c3f6bde37b99cee8e088773793438e910a0cc437cab759a18b2b6e2
+size 10008841510

output/checkpoint-30000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01a525adca3ed55a55b0e32f3565e03eb3de12c488d72632c54c9405dd858a28
+size 14344

output/checkpoint-30000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f4794b5f9540306c90a772b3b6a73944ea7ab02d45c760ac7be8ea8a7e65d86
+size 988

output/checkpoint-30000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:673ee106494e80bd7254452368b4a8a9ef1ab34af463e19284b555963ef6efb4
+size 1000

output/checkpoint-40000/controlnet/config.json

@@ -0,0 +1,57 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.32.0.dev0",
+  "act_fn": "silu",
+  "addition_embed_type": "text_time",
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": 256,
+  "attention_head_dim": [
+    5,
+    10,
+    20
+  ],
+  "block_out_channels": [
+    320,
+    640,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 2048,
+  "down_block_types": [
+    "DownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": 2816,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": [
+    1,
+    2,
+    10
+  ],
+  "upcast_attention": null,
+  "use_linear_projection": true
+}

output/checkpoint-40000/controlnet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d5192eb2d2a16f448e6d9f2f4cf0bba9e70f5ff82f704cfb4f3c5fafb709dee
+size 5004167864

output/checkpoint-40000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fb5f26dd982c7dbd0d9c6d0b34e90dde235380560292773f127f6a49fd19d54
+size 10008841510

output/checkpoint-40000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d644c26fba3d3cb700c68307ca8cb9ac04f54d82b614f513e80db8806db6812d
+size 14344

output/checkpoint-40000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f2bdba071ad840f3d59eece2f6e711860f58323f128e62b5b857210cb6d9e78
+size 988

output/checkpoint-40000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef8d9e6dd1cc417151cc565c35c30235164cd222d36b119a405afcf1c1b3336a
+size 1000

output/checkpoint-50000/controlnet/config.json

@@ -0,0 +1,57 @@
+{
+  "_class_name": "ControlNetModel",
+  "_diffusers_version": "0.32.0.dev0",
+  "act_fn": "silu",
+  "addition_embed_type": "text_time",
+  "addition_embed_type_num_heads": 64,
+  "addition_time_embed_dim": 256,
+  "attention_head_dim": [
+    5,
+    10,
+    20
+  ],
+  "block_out_channels": [
+    320,
+    640,
+    1280
+  ],
+  "class_embed_type": null,
+  "conditioning_channels": 3,
+  "conditioning_embedding_out_channels": [
+    16,
+    32,
+    96,
+    256
+  ],
+  "controlnet_conditioning_channel_order": "rgb",
+  "cross_attention_dim": 2048,
+  "down_block_types": [
+    "DownBlock2D",
+    "CrossAttnDownBlock2D",
+    "CrossAttnDownBlock2D"
+  ],
+  "downsample_padding": 1,
+  "encoder_hid_dim": null,
+  "encoder_hid_dim_type": null,
+  "flip_sin_to_cos": true,
+  "freq_shift": 0,
+  "global_pool_conditions": false,
+  "in_channels": 4,
+  "layers_per_block": 2,
+  "mid_block_scale_factor": 1,
+  "mid_block_type": "UNetMidBlock2DCrossAttn",
+  "norm_eps": 1e-05,
+  "norm_num_groups": 32,
+  "num_attention_heads": null,
+  "num_class_embeds": null,
+  "only_cross_attention": false,
+  "projection_class_embeddings_input_dim": 2816,
+  "resnet_time_scale_shift": "default",
+  "transformer_layers_per_block": [
+    1,
+    2,
+    10
+  ],
+  "upcast_attention": null,
+  "use_linear_projection": true
+}

output/checkpoint-50000/controlnet/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:697916c805339c10cc915c99b64574ade0d02f078cc4b07791bc462b36eba503
+size 5004167864

output/checkpoint-50000/optimizer.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccd8799b95d5519fa9b7d67cd3dac4546375e588b88e83642d57beb264889107
+size 10008841510

output/checkpoint-50000/random_states_0.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4379e91ff6e90233f72224afc7dbf7de2293cb359d55e6f013b1161b32e86f6d
+size 14344

output/checkpoint-50000/scaler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8492ca5c6199ddead96f76b53048d3fbbbc4e32cd1d4f41abe48be425993b1fb
+size 988

output/checkpoint-50000/scheduler.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:884c2374c79400966c8ad3459e73aad51c02318b552c4ffc9fd1792596d58237
+size 1000

output/logs/fill50k_custom_v1_classification_v2/1732146579.9035878/events.out.tfevents.1732146579.f41554fe6d06.14666.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e34905b217b6bc4510060179e060e5bbaa215b51d6fcf0084c52fb677e422ae3
+size 2702

output/logs/fill50k_custom_v1_classification_v2/1732146579.9052796/hparams.yml

@@ -0,0 +1,54 @@
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+caption_column: text
+checkpointing_steps: 10000
+checkpoints_total_limit: null
+conditioning_image_column: conditioning_image
+controlnet_model_name_or_path: null
+crops_coords_top_left_h: 0
+crops_coords_top_left_w: 0
+dataloader_num_workers: 0
+dataset_config_name: null
+dataset_name: null
+enable_npu_flash_attention: false
+enable_xformers_memory_efficient_attention: true
+gradient_accumulation_steps: 1
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+image_column: image
+learning_rate: 1.0e-05
+logging_dir: logs
+lr_num_cycles: 1
+lr_power: 1.0
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 100000
+mixed_precision: fp16
+num_train_epochs: 2
+num_validation_images: 1
+output_dir: output
+pretrained_model_name_or_path: stabilityai/stable-diffusion-xl-base-1.0
+pretrained_vae_model_name_or_path: madebyollin/sdxl-vae-fp16-fix
+proportion_empty_prompts: 0
+push_to_hub: false
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: 43
+set_grads_to_none: false
+tokenizer_name: null
+tracker_project_name: fill50k_custom_v1_classification_v2
+train_batch_size: 1
+train_data_dir: yeq6x/fill50k_custom
+use_8bit_adam: false
+validation_steps: 1000
+variant: null

output/logs/fill50k_custom_v1_classification_v2/1732147449.7879949/events.out.tfevents.1732147449.f41554fe6d06.19394.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:decb931a69bff1fb816694fb00b74ad9cdf3817240a934912f4c9c781889561a
+size 2702

output/logs/fill50k_custom_v1_classification_v2/1732147449.7897315/hparams.yml

@@ -0,0 +1,54 @@
+adam_beta1: 0.9
+adam_beta2: 0.999
+adam_epsilon: 1.0e-08
+adam_weight_decay: 0.01
+allow_tf32: false
+cache_dir: null
+caption_column: text
+checkpointing_steps: 10000
+checkpoints_total_limit: null
+conditioning_image_column: conditioning_image
+controlnet_model_name_or_path: null
+crops_coords_top_left_h: 0
+crops_coords_top_left_w: 0
+dataloader_num_workers: 0
+dataset_config_name: null
+dataset_name: null
+enable_npu_flash_attention: false
+enable_xformers_memory_efficient_attention: true
+gradient_accumulation_steps: 1
+gradient_checkpointing: false
+hub_model_id: null
+hub_token: null
+image_column: image
+learning_rate: 1.0e-05
+logging_dir: logs
+lr_num_cycles: 1
+lr_power: 1.0
+lr_scheduler: constant
+lr_warmup_steps: 500
+max_grad_norm: 1.0
+max_train_samples: null
+max_train_steps: 100000
+mixed_precision: fp16
+num_train_epochs: 2
+num_validation_images: 1
+output_dir: output
+pretrained_model_name_or_path: stabilityai/stable-diffusion-xl-base-1.0
+pretrained_vae_model_name_or_path: madebyollin/sdxl-vae-fp16-fix
+proportion_empty_prompts: 0
+push_to_hub: false
+report_to: tensorboard
+resolution: 512
+resume_from_checkpoint: null
+revision: null
+scale_lr: false
+seed: 43
+set_grads_to_none: false
+tokenizer_name: null
+tracker_project_name: fill50k_custom_v1_classification_v2
+train_batch_size: 1
+train_data_dir: yeq6x/fill50k_custom
+use_8bit_adam: false
+validation_steps: 1000
+variant: null

output/logs/fill50k_custom_v1_classification_v2/events.out.tfevents.1732146579.f41554fe6d06.14666.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94a3ea9d98ade653492259433559d4c345e8cc64abedbf58e62ee45475b35c18
+size 145561

output/logs/fill50k_custom_v1_classification_v2/events.out.tfevents.1732147449.f41554fe6d06.19394.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:933390ca81864cb6ac5f30e6a6d48f8f936bfcca025e57688f59c50802754a39
+size 28954054

output/train_controlnet_sdxl.py

@@ -0,0 +1,1404 @@
+#!/usr/bin/env python
+# coding=utf-8
+# Copyright 2024 The HuggingFace Inc. team. All rights reserved.
+#
+# 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
+
+import argparse
+import functools
+import gc
+import logging
+import math
+import os
+import random
+import shutil
+from contextlib import nullcontext
+from pathlib import Path
+
+import accelerate
+import numpy as np
+import torch
+import torch.nn.functional as F
+import torch.utils.checkpoint
+import transformers
+from accelerate import Accelerator
+from accelerate.logging import get_logger
+from accelerate.utils import DistributedType, ProjectConfiguration, set_seed
+from datasets import load_dataset
+from huggingface_hub import create_repo, upload_folder
+from packaging import version
+from PIL import Image
+from torchvision import transforms
+from tqdm.auto import tqdm
+from transformers import AutoTokenizer, PretrainedConfig
+
+import diffusers
+from diffusers import (
+    AutoencoderKL,
+    ControlNetModel,
+    DDPMScheduler,
+    StableDiffusionXLControlNetPipeline,
+    UNet2DConditionModel,
+    UniPCMultistepScheduler,
+)
+from diffusers.optimization import get_scheduler
+from diffusers.utils import check_min_version, is_wandb_available, make_image_grid
+from diffusers.utils.hub_utils import load_or_create_model_card, populate_model_card
+from diffusers.utils.import_utils import is_torch_npu_available, is_xformers_available
+from diffusers.utils.torch_utils import is_compiled_module
+
+
+if is_wandb_available():
+    import wandb
+
+# Will error if the minimal version of diffusers is not installed. Remove at your own risks.
+check_min_version("0.32.0.dev0")
+
+logger = get_logger(__name__)
+if is_torch_npu_available():
+    torch.npu.config.allow_internal_format = False
+
+
+def log_validation(vae, unet, controlnet, args, accelerator, weight_dtype, step, is_final_validation=False):
+    logger.info("Running validation... ")
+
+    if not is_final_validation:
+        controlnet = accelerator.unwrap_model(controlnet)
+        pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            vae=vae,
+            unet=unet,
+            controlnet=controlnet,
+            revision=args.revision,
+            variant=args.variant,
+            torch_dtype=weight_dtype,
+        )
+    else:
+        controlnet = ControlNetModel.from_pretrained(args.output_dir, torch_dtype=weight_dtype)
+        if args.pretrained_vae_model_name_or_path is not None:
+            vae = AutoencoderKL.from_pretrained(args.pretrained_vae_model_name_or_path, torch_dtype=weight_dtype)
+        else:
+            vae = AutoencoderKL.from_pretrained(
+                args.pretrained_model_name_or_path, subfolder="vae", torch_dtype=weight_dtype
+            )
+
+        pipeline = StableDiffusionXLControlNetPipeline.from_pretrained(
+            args.pretrained_model_name_or_path,
+            vae=vae,
+            controlnet=controlnet,
+            revision=args.revision,
+            variant=args.variant,
+            torch_dtype=weight_dtype,
+        )
+
+    pipeline.scheduler = UniPCMultistepScheduler.from_config(pipeline.scheduler.config)
+    pipeline = pipeline.to(accelerator.device)
+    pipeline.set_progress_bar_config(disable=True)
+
+    if args.enable_xformers_memory_efficient_attention:
+        pipeline.enable_xformers_memory_efficient_attention()
+
+    if args.seed is None:
+        generator = None
+    else:
+        generator = torch.Generator(device=accelerator.device).manual_seed(args.seed)
+
+    if len(args.validation_image) == len(args.validation_prompt):
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_image) == 1:
+        validation_images = args.validation_image * len(args.validation_prompt)
+        validation_prompts = args.validation_prompt
+    elif len(args.validation_prompt) == 1:
+        validation_images = args.validation_image
+        validation_prompts = args.validation_prompt * len(args.validation_image)
+    else:
+        raise ValueError(
+            "number of `args.validation_image` and `args.validation_prompt` should be checked in `parse_args`"
+        )
+
+    image_logs = []
+    if is_final_validation or torch.backends.mps.is_available():
+        autocast_ctx = nullcontext()
+    else:
+        autocast_ctx = torch.autocast(accelerator.device.type)
+
+    for validation_prompt, validation_image in zip(validation_prompts, validation_images):
+        validation_image = Image.open(validation_image).convert("RGB")
+        validation_image = validation_image.resize((args.resolution, args.resolution))
+
+        images = []
+
+        for _ in range(args.num_validation_images):
+            with autocast_ctx:
+                image = pipeline(
+                    prompt=validation_prompt, image=validation_image, num_inference_steps=20, generator=generator
+                ).images[0]
+            images.append(image)
+
+        image_logs.append(
+            {"validation_image": validation_image, "images": images, "validation_prompt": validation_prompt}
+        )
+
+    tracker_key = "test" if is_final_validation else "validation"
+    for tracker in accelerator.trackers:
+        if tracker.name == "tensorboard":
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images = []
+
+                formatted_images.append(np.asarray(validation_image))
+
+                for image in images:
+                    formatted_images.append(np.asarray(image))
+
+                formatted_images = np.stack(formatted_images)
+
+                tracker.writer.add_images(validation_prompt, formatted_images, step, dataformats="NHWC")
+        elif tracker.name == "wandb":
+            formatted_images = []
+
+            for log in image_logs:
+                images = log["images"]
+                validation_prompt = log["validation_prompt"]
+                validation_image = log["validation_image"]
+
+                formatted_images.append(wandb.Image(validation_image, caption="Controlnet conditioning"))
+
+                for image in images:
+                    image = wandb.Image(image, caption=validation_prompt)
+                    formatted_images.append(image)
+
+            tracker.log({tracker_key: formatted_images})
+        else:
+            logger.warning(f"image logging not implemented for {tracker.name}")
+
+        del pipeline
+        gc.collect()
+        torch.cuda.empty_cache()
+
+        return image_logs
+
+
+def import_model_class_from_model_name_or_path(
+    pretrained_model_name_or_path: str, revision: str, subfolder: str = "text_encoder"
+):
+    text_encoder_config = PretrainedConfig.from_pretrained(
+        pretrained_model_name_or_path, subfolder=subfolder, revision=revision
+    )
+    model_class = text_encoder_config.architectures[0]
+
+    if model_class == "CLIPTextModel":
+        from transformers import CLIPTextModel
+
+        return CLIPTextModel
+    elif model_class == "CLIPTextModelWithProjection":
+        from transformers import CLIPTextModelWithProjection
+
+        return CLIPTextModelWithProjection
+    else:
+        raise ValueError(f"{model_class} is not supported.")
+
+
+def save_model_card(repo_id: str, image_logs=None, base_model=str, repo_folder=None):
+    img_str = ""
+    if image_logs is not None:
+        img_str = "You can find some example images below.\n\n"
+        for i, log in enumerate(image_logs):
+            images = log["images"]
+            validation_prompt = log["validation_prompt"]
+            validation_image = log["validation_image"]
+            validation_image.save(os.path.join(repo_folder, "image_control.png"))
+            img_str += f"prompt: {validation_prompt}\n"
+            images = [validation_image] + images
+            make_image_grid(images, 1, len(images)).save(os.path.join(repo_folder, f"images_{i}.png"))
+            img_str += f"![images_{i})](./images_{i}.png)\n"
+
+    model_description = f"""
+# controlnet-{repo_id}
+
+These are controlnet weights trained on {base_model} with new type of conditioning.
+{img_str}
+"""
+
+    model_card = load_or_create_model_card(
+        repo_id_or_path=repo_id,
+        from_training=True,
+        license="openrail++",
+        base_model=base_model,
+        model_description=model_description,
+        inference=True,
+    )
+
+    tags = [
+        "stable-diffusion-xl",
+        "stable-diffusion-xl-diffusers",
+        "text-to-image",
+        "diffusers",
+        "controlnet",
+        "diffusers-training",
+    ]
+    model_card = populate_model_card(model_card, tags=tags)
+
+    model_card.save(os.path.join(repo_folder, "README.md"))
+
+
+def parse_args(input_args=None):
+    parser = argparse.ArgumentParser(description="Simple example of a ControlNet training script.")
+    parser.add_argument(
+        "--pretrained_model_name_or_path",
+        type=str,
+        default=None,
+        required=True,
+        help="Path to pretrained model or model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--pretrained_vae_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to an improved VAE to stabilize training. For more details check out: https://github.com/huggingface/diffusers/pull/4038.",
+    )
+    parser.add_argument(
+        "--controlnet_model_name_or_path",
+        type=str,
+        default=None,
+        help="Path to pretrained controlnet model or model identifier from huggingface.co/models."
+        " If not specified controlnet weights are initialized from unet.",
+    )
+    parser.add_argument(
+        "--variant",
+        type=str,
+        default=None,
+        help="Variant of the model files of the pretrained model identifier from huggingface.co/models, 'e.g.' fp16",
+    )
+    parser.add_argument(
+        "--revision",
+        type=str,
+        default=None,
+        required=False,
+        help="Revision of pretrained model identifier from huggingface.co/models.",
+    )
+    parser.add_argument(
+        "--tokenizer_name",
+        type=str,
+        default=None,
+        help="Pretrained tokenizer name or path if not the same as model_name",
+    )
+    parser.add_argument(
+        "--output_dir",
+        type=str,
+        default="controlnet-model",
+        help="The output directory where the model predictions and checkpoints will be written.",
+    )
+    parser.add_argument(
+        "--cache_dir",
+        type=str,
+        default=None,
+        help="The directory where the downloaded models and datasets will be stored.",
+    )
+    parser.add_argument("--seed", type=int, default=None, help="A seed for reproducible training.")
+    parser.add_argument(
+        "--resolution",
+        type=int,
+        default=512,
+        help=(
+            "The resolution for input images, all the images in the train/validation dataset will be resized to this"
+            " resolution"
+        ),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_h",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--crops_coords_top_left_w",
+        type=int,
+        default=0,
+        help=("Coordinate for (the height) to be included in the crop coordinate embeddings needed by SDXL UNet."),
+    )
+    parser.add_argument(
+        "--train_batch_size", type=int, default=4, help="Batch size (per device) for the training dataloader."
+    )
+    parser.add_argument("--num_train_epochs", type=int, default=1)
+    parser.add_argument(
+        "--max_train_steps",
+        type=int,
+        default=None,
+        help="Total number of training steps to perform.  If provided, overrides num_train_epochs.",
+    )
+    parser.add_argument(
+        "--checkpointing_steps",
+        type=int,
+        default=500,
+        help=(
+            "Save a checkpoint of the training state every X updates. Checkpoints can be used for resuming training via `--resume_from_checkpoint`. "
+            "In the case that the checkpoint is better than the final trained model, the checkpoint can also be used for inference."
+            "Using a checkpoint for inference requires separate loading of the original pipeline and the individual checkpointed model components."
+            "See https://huggingface.co/docs/diffusers/main/en/training/dreambooth#performing-inference-using-a-saved-checkpoint for step by step"
+            "instructions."
+        ),
+    )
+    parser.add_argument(
+        "--checkpoints_total_limit",
+        type=int,
+        default=None,
+        help=("Max number of checkpoints to store."),
+    )
+    parser.add_argument(
+        "--resume_from_checkpoint",
+        type=str,
+        default=None,
+        help=(
+            "Whether training should be resumed from a previous checkpoint. Use a path saved by"
+            ' `--checkpointing_steps`, or `"latest"` to automatically select the last available checkpoint.'
+        ),
+    )
+    parser.add_argument(
+        "--gradient_accumulation_steps",
+        type=int,
+        default=1,
+        help="Number of updates steps to accumulate before performing a backward/update pass.",
+    )
+    parser.add_argument(
+        "--gradient_checkpointing",
+        action="store_true",
+        help="Whether or not to use gradient checkpointing to save memory at the expense of slower backward pass.",
+    )
+    parser.add_argument(
+        "--learning_rate",
+        type=float,
+        default=5e-6,
+        help="Initial learning rate (after the potential warmup period) to use.",
+    )
+    parser.add_argument(
+        "--scale_lr",
+        action="store_true",
+        default=False,
+        help="Scale the learning rate by the number of GPUs, gradient accumulation steps, and batch size.",
+    )
+    parser.add_argument(
+        "--lr_scheduler",
+        type=str,
+        default="constant",
+        help=(
+            'The scheduler type to use. Choose between ["linear", "cosine", "cosine_with_restarts", "polynomial",'
+            ' "constant", "constant_with_warmup"]'
+        ),
+    )
+    parser.add_argument(
+        "--lr_warmup_steps", type=int, default=500, help="Number of steps for the warmup in the lr scheduler."
+    )
+    parser.add_argument(
+        "--lr_num_cycles",
+        type=int,
+        default=1,
+        help="Number of hard resets of the lr in cosine_with_restarts scheduler.",
+    )
+    parser.add_argument("--lr_power", type=float, default=1.0, help="Power factor of the polynomial scheduler.")
+    parser.add_argument(
+        "--use_8bit_adam", action="store_true", help="Whether or not to use 8-bit Adam from bitsandbytes."
+    )
+    parser.add_argument(
+        "--dataloader_num_workers",
+        type=int,
+        default=0,
+        help=(
+            "Number of subprocesses to use for data loading. 0 means that the data will be loaded in the main process."
+        ),
+    )
+    parser.add_argument("--adam_beta1", type=float, default=0.9, help="The beta1 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_beta2", type=float, default=0.999, help="The beta2 parameter for the Adam optimizer.")
+    parser.add_argument("--adam_weight_decay", type=float, default=1e-2, help="Weight decay to use.")
+    parser.add_argument("--adam_epsilon", type=float, default=1e-08, help="Epsilon value for the Adam optimizer")
+    parser.add_argument("--max_grad_norm", default=1.0, type=float, help="Max gradient norm.")
+    parser.add_argument("--push_to_hub", action="store_true", help="Whether or not to push the model to the Hub.")
+    parser.add_argument("--hub_token", type=str, default=None, help="The token to use to push to the Model Hub.")
+    parser.add_argument(
+        "--hub_model_id",
+        type=str,
+        default=None,
+        help="The name of the repository to keep in sync with the local `output_dir`.",
+    )
+    parser.add_argument(
+        "--logging_dir",
+        type=str,
+        default="logs",
+        help=(
+            "[TensorBoard](https://www.tensorflow.org/tensorboard) log directory. Will default to"
+            " *output_dir/runs/**CURRENT_DATETIME_HOSTNAME***."
+        ),
+    )
+    parser.add_argument(
+        "--allow_tf32",
+        action="store_true",
+        help=(
+            "Whether or not to allow TF32 on Ampere GPUs. Can be used to speed up training. For more information, see"
+            " https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices"
+        ),
+    )
+    parser.add_argument(
+        "--report_to",
+        type=str,
+        default="tensorboard",
+        help=(
+            'The integration to report the results and logs to. Supported platforms are `"tensorboard"`'
+            ' (default), `"wandb"` and `"comet_ml"`. Use `"all"` to report to all integrations.'
+        ),
+    )
+    parser.add_argument(
+        "--mixed_precision",
+        type=str,
+        default=None,
+        choices=["no", "fp16", "bf16"],
+        help=(
+            "Whether to use mixed precision. Choose between fp16 and bf16 (bfloat16). Bf16 requires PyTorch >="
+            " 1.10.and an Nvidia Ampere GPU.  Default to the value of accelerate config of the current system or the"
+            " flag passed with the `accelerate.launch` command. Use this argument to override the accelerate config."
+        ),
+    )
+    parser.add_argument(
+        "--enable_xformers_memory_efficient_attention", action="store_true", help="Whether or not to use xformers."
+    )
+    parser.add_argument(
+        "--enable_npu_flash_attention", action="store_true", help="Whether or not to use npu flash attention."
+    )
+    parser.add_argument(
+        "--set_grads_to_none",
+        action="store_true",
+        help=(
+            "Save more memory by using setting grads to None instead of zero. Be aware, that this changes certain"
+            " behaviors, so disable this argument if it causes any problems. More info:"
+            " https://pytorch.org/docs/stable/generated/torch.optim.Optimizer.zero_grad.html"
+        ),
+    )
+    parser.add_argument(
+        "--dataset_name",
+        type=str,
+        default=None,
+        help=(
+            "The name of the Dataset (from the HuggingFace hub) to train on (could be your own, possibly private,"
+            " dataset). It can also be a path pointing to a local copy of a dataset in your filesystem,"
+            " or to a folder containing files that 🤗 Datasets can understand."
+        ),
+    )
+    parser.add_argument(
+        "--dataset_config_name",
+        type=str,
+        default=None,
+        help="The config of the Dataset, leave as None if there's only one config.",
+    )
+    parser.add_argument(
+        "--train_data_dir",
+        type=str,
+        default=None,
+        help=(
+            "A folder containing the training data. Folder contents must follow the structure described in"
+            " https://huggingface.co/docs/datasets/image_dataset#imagefolder. In particular, a `metadata.jsonl` file"
+            " must exist to provide the captions for the images. Ignored if `dataset_name` is specified."
+        ),
+    )
+    parser.add_argument(
+        "--image_column", type=str, default="image", help="The column of the dataset containing the target image."
+    )
+    parser.add_argument(
+        "--conditioning_image_column",
+        type=str,
+        default="conditioning_image",
+        help="The column of the dataset containing the controlnet conditioning image.",
+    )
+    parser.add_argument(
+        "--caption_column",
+        type=str,
+        default="text",
+        help="The column of the dataset containing a caption or a list of captions.",
+    )
+    parser.add_argument(
+        "--max_train_samples",
+        type=int,
+        default=None,
+        help=(
+            "For debugging purposes or quicker training, truncate the number of training examples to this "
+            "value if set."
+        ),
+    )
+    parser.add_argument(
+        "--proportion_empty_prompts",
+        type=float,
+        default=0,
+        help="Proportion of image prompts to be replaced with empty strings. Defaults to 0 (no prompt replacement).",
+    )
+    parser.add_argument(
+        "--validation_prompt",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of prompts evaluated every `--validation_steps` and logged to `--report_to`."
+            " Provide either a matching number of `--validation_image`s, a single `--validation_image`"
+            " to be used with all prompts, or a single prompt that will be used with all `--validation_image`s."
+        ),
+    )
+    parser.add_argument(
+        "--validation_image",
+        type=str,
+        default=None,
+        nargs="+",
+        help=(
+            "A set of paths to the controlnet conditioning image be evaluated every `--validation_steps`"
+            " and logged to `--report_to`. Provide either a matching number of `--validation_prompt`s, a"
+            " a single `--validation_prompt` to be used with all `--validation_image`s, or a single"
+            " `--validation_image` that will be used with all `--validation_prompt`s."
+        ),
+    )
+    parser.add_argument(
+        "--num_validation_images",
+        type=int,
+        default=4,
+        help="Number of images to be generated for each `--validation_image`, `--validation_prompt` pair",
+    )
+    parser.add_argument(
+        "--validation_steps",
+        type=int,
+        default=100,
+        help=(
+            "Run validation every X steps. Validation consists of running the prompt"
+            " `args.validation_prompt` multiple times: `args.num_validation_images`"
+            " and logging the images."
+        ),
+    )
+    parser.add_argument(
+        "--tracker_project_name",
+        type=str,
+        default="sd_xl_train_controlnet",
+        help=(
+            "The `project_name` argument passed to Accelerator.init_trackers for"
+            " more information see https://huggingface.co/docs/accelerate/v0.17.0/en/package_reference/accelerator#accelerate.Accelerator"
+        ),
+    )
+
+    if input_args is not None:
+        args = parser.parse_args(input_args)
+    else:
+        args = parser.parse_args()
+
+    if args.dataset_name is None and args.train_data_dir is None:
+        raise ValueError("Specify either `--dataset_name` or `--train_data_dir`")
+
+    if args.dataset_name is not None and args.train_data_dir is not None:
+        raise ValueError("Specify only one of `--dataset_name` or `--train_data_dir`")
+
+    if args.proportion_empty_prompts < 0 or args.proportion_empty_prompts > 1:
+        raise ValueError("`--proportion_empty_prompts` must be in the range [0, 1].")
+
+    if args.validation_prompt is not None and args.validation_image is None:
+        raise ValueError("`--validation_image` must be set if `--validation_prompt` is set")
+
+    if args.validation_prompt is None and args.validation_image is not None:
+        raise ValueError("`--validation_prompt` must be set if `--validation_image` is set")
+
+    if (
+        args.validation_image is not None
+        and args.validation_prompt is not None
+        and len(args.validation_image) != 1
+        and len(args.validation_prompt) != 1
+        and len(args.validation_image) != len(args.validation_prompt)
+    ):
+        raise ValueError(
+            "Must provide either 1 `--validation_image`, 1 `--validation_prompt`,"
+            " or the same number of `--validation_prompt`s and `--validation_image`s"
+        )
+
+    if args.resolution % 8 != 0:
+        raise ValueError(
+            "`--resolution` must be divisible by 8 for consistently sized encoded images between the VAE and the controlnet encoder."
+        )
+
+    return args
+
+
+def get_train_dataset(args, accelerator):
+    # Get the datasets: you can either provide your own training and evaluation files (see below)
+    # or specify a Dataset from the hub (the dataset will be downloaded automatically from the datasets Hub).
+
+    # In distributed training, the load_dataset function guarantees that only one local process can concurrently
+    # download the dataset.
+    if args.dataset_name is not None:
+        # Downloading and loading a dataset from the hub.
+        dataset = load_dataset(
+            args.dataset_name,
+            args.dataset_config_name,
+            cache_dir=args.cache_dir,
+        )
+    else:
+        if args.train_data_dir is not None:
+            dataset = load_dataset(
+                args.train_data_dir,
+                cache_dir=args.cache_dir,
+            )
+        # See more about loading custom images at
+        # https://huggingface.co/docs/datasets/v2.0.0/en/dataset_script
+
+    # Preprocessing the datasets.
+    # We need to tokenize inputs and targets.
+    column_names = dataset["train"].column_names
+
+    # 6. Get the column names for input/target.
+    if args.image_column is None:
+        image_column = column_names[0]
+        logger.info(f"image column defaulting to {image_column}")
+    else:
+        image_column = args.image_column
+        if image_column not in column_names:
+            raise ValueError(
+                f"`--image_column` value '{args.image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.caption_column is None:
+        caption_column = column_names[1]
+        logger.info(f"caption column defaulting to {caption_column}")
+    else:
+        caption_column = args.caption_column
+        if caption_column not in column_names:
+            raise ValueError(
+                f"`--caption_column` value '{args.caption_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    if args.conditioning_image_column is None:
+        conditioning_image_column = column_names[2]
+        logger.info(f"conditioning image column defaulting to {conditioning_image_column}")
+    else:
+        conditioning_image_column = args.conditioning_image_column
+        if conditioning_image_column not in column_names:
+            raise ValueError(
+                f"`--conditioning_image_column` value '{args.conditioning_image_column}' not found in dataset columns. Dataset columns are: {', '.join(column_names)}"
+            )
+
+    with accelerator.main_process_first():
+        train_dataset = dataset["train"].shuffle(seed=args.seed)
+        if args.max_train_samples is not None:
+            train_dataset = train_dataset.select(range(args.max_train_samples))
+    return train_dataset
+
+
+# Adapted from pipelines.StableDiffusionXLPipeline.encode_prompt
+def encode_prompt(prompt_batch, text_encoders, tokenizers, proportion_empty_prompts, is_train=True):
+    prompt_embeds_list = []
+
+    captions = []
+    for caption in prompt_batch:
+        if random.random() < proportion_empty_prompts:
+            captions.append("")
+        elif isinstance(caption, str):
+            captions.append(caption)
+        elif isinstance(caption, (list, np.ndarray)):
+            # take a random caption if there are multiple
+            captions.append(random.choice(caption) if is_train else caption[0])
+
+    with torch.no_grad():
+        for tokenizer, text_encoder in zip(tokenizers, text_encoders):
+            text_inputs = tokenizer(
+                captions,
+                padding="max_length",
+                max_length=tokenizer.model_max_length,
+                truncation=True,
+                return_tensors="pt",
+            )
+            text_input_ids = text_inputs.input_ids
+            prompt_embeds = text_encoder(
+                text_input_ids.to(text_encoder.device),
+                output_hidden_states=True,
+            )
+
+            # We are only ALWAYS interested in the pooled output of the final text encoder
+            pooled_prompt_embeds = prompt_embeds[0]
+            prompt_embeds = prompt_embeds.hidden_states[-2]
+            bs_embed, seq_len, _ = prompt_embeds.shape
+            prompt_embeds = prompt_embeds.view(bs_embed, seq_len, -1)
+            prompt_embeds_list.append(prompt_embeds)
+
+    prompt_embeds = torch.concat(prompt_embeds_list, dim=-1)
+    pooled_prompt_embeds = pooled_prompt_embeds.view(bs_embed, -1)
+    return prompt_embeds, pooled_prompt_embeds
+
+
+def prepare_train_dataset(dataset, accelerator):
+    image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+            transforms.Normalize([0.5], [0.5]),
+        ]
+    )
+
+    conditioning_image_transforms = transforms.Compose(
+        [
+            transforms.Resize(args.resolution, interpolation=transforms.InterpolationMode.BILINEAR),
+            transforms.CenterCrop(args.resolution),
+            transforms.ToTensor(),
+        ]
+    )
+
+    def preprocess_train(examples):
+        images = [image.convert("RGB") for image in examples[args.image_column]]
+        images = [image_transforms(image) for image in images]
+
+        conditioning_images = [image.convert("RGB") for image in examples[args.conditioning_image_column]]
+        conditioning_images = [conditioning_image_transforms(image) for image in conditioning_images]
+
+        examples["pixel_values"] = images
+        examples["conditioning_pixel_values"] = conditioning_images
+
+        return examples
+
+    with accelerator.main_process_first():
+        dataset = dataset.with_transform(preprocess_train)
+
+    return dataset
+
+
+def collate_fn(examples):
+    pixel_values = torch.stack([example["pixel_values"] for example in examples])
+    pixel_values = pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    conditioning_pixel_values = torch.stack([example["conditioning_pixel_values"] for example in examples])
+    conditioning_pixel_values = conditioning_pixel_values.to(memory_format=torch.contiguous_format).float()
+
+    prompt_ids = torch.stack([torch.tensor(example["prompt_embeds"]) for example in examples])
+
+    add_text_embeds = torch.stack([torch.tensor(example["text_embeds"]) for example in examples])
+    add_time_ids = torch.stack([torch.tensor(example["time_ids"]) for example in examples])
+
+    return {
+        "pixel_values": pixel_values,
+        "conditioning_pixel_values": conditioning_pixel_values,
+        "prompt_ids": prompt_ids,
+        "unet_added_conditions": {"text_embeds": add_text_embeds, "time_ids": add_time_ids},
+        "text": [example["text"] for example in examples]
+    }
+
+
+def main(args):
+    if args.report_to == "wandb" and args.hub_token is not None:
+        raise ValueError(
+            "You cannot use both --report_to=wandb and --hub_token due to a security risk of exposing your token."
+            " Please use `huggingface-cli login` to authenticate with the Hub."
+        )
+
+    logging_dir = Path(args.output_dir, args.logging_dir)
+
+    if torch.backends.mps.is_available() and args.mixed_precision == "bf16":
+        # due to pytorch#99272, MPS does not yet support bfloat16.
+        raise ValueError(
+            "Mixed precision training with bfloat16 is not supported on MPS. Please use fp16 (recommended) or fp32 instead."
+        )
+
+    accelerator_project_config = ProjectConfiguration(project_dir=args.output_dir, logging_dir=logging_dir)
+
+    accelerator = Accelerator(
+        gradient_accumulation_steps=args.gradient_accumulation_steps,
+        mixed_precision=args.mixed_precision,
+        log_with=args.report_to,
+        project_config=accelerator_project_config,
+    )
+
+    # Disable AMP for MPS.
+    if torch.backends.mps.is_available():
+        accelerator.native_amp = False
+
+    # Make one log on every process with the configuration for debugging.
+    logging.basicConfig(
+        format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
+        datefmt="%m/%d/%Y %H:%M:%S",
+        level=logging.INFO,
+    )
+    logger.info(accelerator.state, main_process_only=False)
+    if accelerator.is_local_main_process:
+        transformers.utils.logging.set_verbosity_warning()
+        diffusers.utils.logging.set_verbosity_info()
+    else:
+        transformers.utils.logging.set_verbosity_error()
+        diffusers.utils.logging.set_verbosity_error()
+
+    # If passed along, set the training seed now.
+    if args.seed is not None:
+        set_seed(args.seed)
+
+    # Handle the repository creation
+    if accelerator.is_main_process:
+        if args.output_dir is not None:
+            os.makedirs(args.output_dir, exist_ok=True)
+
+        if args.push_to_hub:
+            repo_id = create_repo(
+                repo_id=args.hub_model_id or Path(args.output_dir).name, exist_ok=True, token=args.hub_token
+            ).repo_id
+
+    # Load the tokenizers
+    tokenizer_one = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer",
+        revision=args.revision,
+        use_fast=False,
+    )
+    tokenizer_two = AutoTokenizer.from_pretrained(
+        args.pretrained_model_name_or_path,
+        subfolder="tokenizer_2",
+        revision=args.revision,
+        use_fast=False,
+    )
+
+    # import correct text encoder classes
+    text_encoder_cls_one = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision
+    )
+    text_encoder_cls_two = import_model_class_from_model_name_or_path(
+        args.pretrained_model_name_or_path, args.revision, subfolder="text_encoder_2"
+    )
+
+    # Load scheduler and models
+    noise_scheduler = DDPMScheduler.from_pretrained(args.pretrained_model_name_or_path, subfolder="scheduler")
+    text_encoder_one = text_encoder_cls_one.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder", revision=args.revision, variant=args.variant
+    )
+    text_encoder_two = text_encoder_cls_two.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="text_encoder_2", revision=args.revision, variant=args.variant
+    )
+    vae_path = (
+        args.pretrained_model_name_or_path
+        if args.pretrained_vae_model_name_or_path is None
+        else args.pretrained_vae_model_name_or_path
+    )
+    vae = AutoencoderKL.from_pretrained(
+        vae_path,
+        subfolder="vae" if args.pretrained_vae_model_name_or_path is None else None,
+        revision=args.revision,
+        variant=args.variant,
+    )
+    unet = UNet2DConditionModel.from_pretrained(
+        args.pretrained_model_name_or_path, subfolder="unet", revision=args.revision, variant=args.variant
+    )
+
+    if args.controlnet_model_name_or_path:
+        logger.info("Loading existing controlnet weights")
+        controlnet = ControlNetModel.from_pretrained(args.controlnet_model_name_or_path)
+    else:
+        logger.info("Initializing controlnet weights from unet")
+        controlnet = ControlNetModel.from_unet(unet)
+
+    def unwrap_model(model):
+        model = accelerator.unwrap_model(model)
+        model = model._orig_mod if is_compiled_module(model) else model
+        return model
+
+    # `accelerate` 0.16.0 will have better support for customized saving
+    if version.parse(accelerate.__version__) >= version.parse("0.16.0"):
+        # create custom saving & loading hooks so that `accelerator.save_state(...)` serializes in a nice format
+        def save_model_hook(models, weights, output_dir):
+            if accelerator.is_main_process:
+                i = len(weights) - 1
+
+                while len(weights) > 0:
+                    weights.pop()
+                    model = models[i]
+
+                    sub_dir = "controlnet"
+                    model.save_pretrained(os.path.join(output_dir, sub_dir))
+
+                    i -= 1
+
+        def load_model_hook(models, input_dir):
+            while len(models) > 0:
+                # pop models so that they are not loaded again
+                model = models.pop()
+
+                # load diffusers style into model
+                load_model = ControlNetModel.from_pretrained(input_dir, subfolder="controlnet")
+                model.register_to_config(**load_model.config)
+
+                model.load_state_dict(load_model.state_dict())
+                del load_model
+
+        accelerator.register_save_state_pre_hook(save_model_hook)
+        accelerator.register_load_state_pre_hook(load_model_hook)
+
+    vae.requires_grad_(False)
+    unet.requires_grad_(False)
+    text_encoder_one.requires_grad_(False)
+    text_encoder_two.requires_grad_(False)
+    controlnet.train()
+
+    if args.enable_npu_flash_attention:
+        if is_torch_npu_available():
+            logger.info("npu flash attention enabled.")
+            unet.enable_npu_flash_attention()
+        else:
+            raise ValueError("npu flash attention requires torch_npu extensions and is supported only on npu devices.")
+
+    if args.enable_xformers_memory_efficient_attention:
+        if is_xformers_available():
+            import xformers
+
+            xformers_version = version.parse(xformers.__version__)
+            if xformers_version == version.parse("0.0.16"):
+                logger.warning(
+                    "xFormers 0.0.16 cannot be used for training in some GPUs. If you observe problems during training, please update xFormers to at least 0.0.17. See https://huggingface.co/docs/diffusers/main/en/optimization/xformers for more details."
+                )
+            unet.enable_xformers_memory_efficient_attention()
+            controlnet.enable_xformers_memory_efficient_attention()
+        else:
+            raise ValueError("xformers is not available. Make sure it is installed correctly")
+
+    if args.gradient_checkpointing:
+        controlnet.enable_gradient_checkpointing()
+        unet.enable_gradient_checkpointing()
+
+    # Check that all trainable models are in full precision
+    low_precision_error_string = (
+        " Please make sure to always have all model weights in full float32 precision when starting training - even if"
+        " doing mixed precision training, copy of the weights should still be float32."
+    )
+
+    if unwrap_model(controlnet).dtype != torch.float32:
+        raise ValueError(
+            f"Controlnet loaded as datatype {unwrap_model(controlnet).dtype}. {low_precision_error_string}"
+        )
+
+    # Enable TF32 for faster training on Ampere GPUs,
+    # cf https://pytorch.org/docs/stable/notes/cuda.html#tensorfloat-32-tf32-on-ampere-devices
+    if args.allow_tf32:
+        torch.backends.cuda.matmul.allow_tf32 = True
+
+    if args.scale_lr:
+        args.learning_rate = (
+            args.learning_rate * args.gradient_accumulation_steps * args.train_batch_size * accelerator.num_processes
+        )
+
+    # Use 8-bit Adam for lower memory usage or to fine-tune the model in 16GB GPUs
+    if args.use_8bit_adam:
+        try:
+            import bitsandbytes as bnb
+        except ImportError:
+            raise ImportError(
+                "To use 8-bit Adam, please install the bitsandbytes library: `pip install bitsandbytes`."
+            )
+
+        optimizer_class = bnb.optim.AdamW8bit
+    else:
+        optimizer_class = torch.optim.AdamW
+
+    # Optimizer creation
+    params_to_optimize = controlnet.parameters()
+    optimizer = optimizer_class(
+        params_to_optimize,
+        lr=args.learning_rate,
+        betas=(args.adam_beta1, args.adam_beta2),
+        weight_decay=args.adam_weight_decay,
+        eps=args.adam_epsilon,
+    )
+
+    # For mixed precision training we cast the text_encoder and vae weights to half-precision
+    # as these models are only used for inference, keeping weights in full precision is not required.
+    weight_dtype = torch.float32
+    if accelerator.mixed_precision == "fp16":
+        weight_dtype = torch.float16
+    elif accelerator.mixed_precision == "bf16":
+        weight_dtype = torch.bfloat16
+
+    # Move vae, unet and text_encoder to device and cast to weight_dtype
+    # The VAE is in float32 to avoid NaN losses.
+    if args.pretrained_vae_model_name_or_path is not None:
+        vae.to(accelerator.device, dtype=weight_dtype)
+    else:
+        vae.to(accelerator.device, dtype=torch.float32)
+    unet.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_one.to(accelerator.device, dtype=weight_dtype)
+    text_encoder_two.to(accelerator.device, dtype=weight_dtype)
+
+    # Here, we compute not just the text embeddings but also the additional embeddings
+    # needed for the SD XL UNet to operate.
+    def compute_embeddings(batch, proportion_empty_prompts, text_encoders, tokenizers, is_train=True):
+        original_size = (args.resolution, args.resolution)
+        target_size = (args.resolution, args.resolution)
+        crops_coords_top_left = (args.crops_coords_top_left_h, args.crops_coords_top_left_w)
+        prompt_batch = batch[args.caption_column]
+
+        prompt_embeds, pooled_prompt_embeds = encode_prompt(
+            prompt_batch, text_encoders, tokenizers, proportion_empty_prompts, is_train
+        )
+        add_text_embeds = pooled_prompt_embeds
+
+        # Adapted from pipeline.StableDiffusionXLPipeline._get_add_time_ids
+        add_time_ids = list(original_size + crops_coords_top_left + target_size)
+        add_time_ids = torch.tensor([add_time_ids])
+
+        prompt_embeds = prompt_embeds.to(accelerator.device)
+        add_text_embeds = add_text_embeds.to(accelerator.device)
+        add_time_ids = add_time_ids.repeat(len(prompt_batch), 1)
+        add_time_ids = add_time_ids.to(accelerator.device, dtype=prompt_embeds.dtype)
+        unet_added_cond_kwargs = {"text_embeds": add_text_embeds, "time_ids": add_time_ids}
+
+        return {"prompt_embeds": prompt_embeds, **unet_added_cond_kwargs}
+
+    # Let's first compute all the embeddings so that we can free up the text encoders
+    # from memory.
+    text_encoders = [text_encoder_one, text_encoder_two]
+    tokenizers = [tokenizer_one, tokenizer_two]
+    train_dataset = get_train_dataset(args, accelerator)
+    compute_embeddings_fn = functools.partial(
+        compute_embeddings,
+        text_encoders=text_encoders,
+        tokenizers=tokenizers,
+        proportion_empty_prompts=args.proportion_empty_prompts,
+    )
+    with accelerator.main_process_first():
+        from datasets.fingerprint import Hasher
+
+        # fingerprint used by the cache for the other processes to load the result
+        # details: https://github.com/huggingface/diffusers/pull/4038#discussion_r1266078401
+        new_fingerprint = Hasher.hash(args)
+        train_dataset = train_dataset.map(compute_embeddings_fn, batched=True, new_fingerprint=new_fingerprint)
+
+    del text_encoders, tokenizers
+    gc.collect()
+    torch.cuda.empty_cache()
+
+    # Then get the training dataset ready to be passed to the dataloader.
+    train_dataset = prepare_train_dataset(train_dataset, accelerator)
+
+    train_dataloader = torch.utils.data.DataLoader(
+        train_dataset,
+        shuffle=True,
+        collate_fn=collate_fn,
+        batch_size=args.train_batch_size,
+        num_workers=args.dataloader_num_workers,
+    )
+
+    # Scheduler and math around the number of training steps.
+    # Check the PR https://github.com/huggingface/diffusers/pull/8312 for detailed explanation.
+    num_warmup_steps_for_scheduler = args.lr_warmup_steps * accelerator.num_processes
+    if args.max_train_steps is None:
+        len_train_dataloader_after_sharding = math.ceil(len(train_dataloader) / accelerator.num_processes)
+        num_update_steps_per_epoch = math.ceil(len_train_dataloader_after_sharding / args.gradient_accumulation_steps)
+        num_training_steps_for_scheduler = (
+            args.num_train_epochs * num_update_steps_per_epoch * accelerator.num_processes
+        )
+    else:
+        num_training_steps_for_scheduler = args.max_train_steps * accelerator.num_processes
+
+    lr_scheduler = get_scheduler(
+        args.lr_scheduler,
+        optimizer=optimizer,
+        num_warmup_steps=num_warmup_steps_for_scheduler,
+        num_training_steps=num_training_steps_for_scheduler,
+        num_cycles=args.lr_num_cycles,
+        power=args.lr_power,
+    )
+
+    # Prepare everything with our `accelerator`.
+    controlnet, optimizer, train_dataloader, lr_scheduler = accelerator.prepare(
+        controlnet, optimizer, train_dataloader, lr_scheduler
+    )
+
+    # We need to recalculate our total training steps as the size of the training dataloader may have changed.
+    num_update_steps_per_epoch = math.ceil(len(train_dataloader) / args.gradient_accumulation_steps)
+    if args.max_train_steps is None:
+        args.max_train_steps = args.num_train_epochs * num_update_steps_per_epoch
+        if num_training_steps_for_scheduler != args.max_train_steps * accelerator.num_processes:
+            logger.warning(
+                f"The length of the 'train_dataloader' after 'accelerator.prepare' ({len(train_dataloader)}) does not match "
+                f"the expected length ({len_train_dataloader_after_sharding}) when the learning rate scheduler was created. "
+                f"This inconsistency may result in the learning rate scheduler not functioning properly."
+            )
+    # Afterwards we recalculate our number of training epochs
+    args.num_train_epochs = math.ceil(args.max_train_steps / num_update_steps_per_epoch)
+
+    # We need to initialize the trackers we use, and also store our configuration.
+    # The trackers initializes automatically on the main process.
+    if accelerator.is_main_process:
+        tracker_config = dict(vars(args))
+
+        # tensorboard cannot handle list types for config
+        tracker_config.pop("validation_prompt")
+        tracker_config.pop("validation_image")
+
+        accelerator.init_trackers(args.tracker_project_name, config=tracker_config)
+
+    # Train!
+    total_batch_size = args.train_batch_size * accelerator.num_processes * args.gradient_accumulation_steps
+
+    logger.info("***** Running training *****")
+    logger.info(f"  Num examples = {len(train_dataset)}")
+    logger.info(f"  Num batches each epoch = {len(train_dataloader)}")
+    logger.info(f"  Num Epochs = {args.num_train_epochs}")
+    logger.info(f"  Instantaneous batch size per device = {args.train_batch_size}")
+    logger.info(f"  Total train batch size (w. parallel, distributed & accumulation) = {total_batch_size}")
+    logger.info(f"  Gradient Accumulation steps = {args.gradient_accumulation_steps}")
+    logger.info(f"  Total optimization steps = {args.max_train_steps}")
+    global_step = 0
+    first_epoch = 0
+
+    # Potentially load in the weights and states from a previous save
+    if args.resume_from_checkpoint:
+        if args.resume_from_checkpoint != "latest":
+            path = os.path.basename(args.resume_from_checkpoint)
+        else:
+            # Get the most recent checkpoint
+            dirs = os.listdir(args.output_dir)
+            dirs = [d for d in dirs if d.startswith("checkpoint")]
+            dirs = sorted(dirs, key=lambda x: int(x.split("-")[1]))
+            path = dirs[-1] if len(dirs) > 0 else None
+
+        if path is None:
+            accelerator.print(
+                f"Checkpoint '{args.resume_from_checkpoint}' does not exist. Starting a new training run."
+            )
+            args.resume_from_checkpoint = None
+            initial_global_step = 0
+        else:
+            accelerator.print(f"Resuming from checkpoint {path}")
+            accelerator.load_state(os.path.join(args.output_dir, path))
+            global_step = int(path.split("-")[1])
+
+            initial_global_step = global_step
+            first_epoch = global_step // num_update_steps_per_epoch
+    else:
+        initial_global_step = 0
+
+    progress_bar = tqdm(
+        range(0, args.max_train_steps),
+        initial=initial_global_step,
+        desc="Steps",
+        # Only show the progress bar once on each machine.
+        disable=not accelerator.is_local_main_process,
+    )
+
+    from torch import nn
+
+    class MultiLayerClassifierHead(nn.Module):
+        def __init__(self, input_dim, hidden_dim, num_classes):
+            super(MultiLayerClassifierHead, self).__init__()
+            self.fc1 = nn.Linear(input_dim, hidden_dim)  # 入力 -> 隠れ層
+            self.fc2 = nn.Linear(hidden_dim, num_classes)  # 隠れ層 -> 出力
+            self.activation = nn.ReLU()  # 非線形活性化関数
+
+        def forward(self, x):
+            # Global Average Pooling
+            x = F.adaptive_avg_pool2d(x, (1, 1))  # [batch, channels, h, w] -> [batch, channels, 1, 1]
+            x = x.view(x.size(0), -1)  # Flatten: [batch, channels]
+            
+            # 2層の全結合層
+            x = self.fc1(x)  # 隠れ層
+            x = self.activation(x)  # 活性化
+            x = self.fc2(x)  # 出力層
+            return x
+
+    # クラスの種類を定義
+    class_types = ["circle", "star", "octagon"]
+
+    # クラス名をインデックスに変換する辞書
+    class_to_idx = {cls: idx for idx, cls in enumerate(class_types)}
+    print(class_to_idx)  # {'circle': 0, 'star': 1, 'octagon': 2}
+
+    # 分類ヘッドの初期化
+    classification_head = MultiLayerClassifierHead(input_dim=1280, hidden_dim=512, num_classes=len(class_types))
+    classification_head = classification_head.to(accelerator.device)
+
+    image_logs = None
+    for epoch in range(first_epoch, args.num_train_epochs):
+        for step, batch in enumerate(train_dataloader):
+            with accelerator.accumulate(controlnet):
+                # Convert images to latent space
+                if args.pretrained_vae_model_name_or_path is not None:
+                    pixel_values = batch["pixel_values"].to(dtype=weight_dtype)
+                else:
+                    pixel_values = batch["pixel_values"]
+                latents = vae.encode(pixel_values).latent_dist.sample()
+                latents = latents * vae.config.scaling_factor
+                if args.pretrained_vae_model_name_or_path is None:
+                    latents = latents.to(weight_dtype)
+
+                # Sample noise that we'll add to the latents
+                noise = torch.randn_like(latents)
+                bsz = latents.shape[0]
+
+                # Sample a random timestep for each image
+                timesteps = torch.randint(0, noise_scheduler.config.num_train_timesteps, (bsz,), device=latents.device)
+                timesteps = timesteps.long()
+
+                # Add noise to the latents according to the noise magnitude at each timestep
+                # (this is the forward diffusion process)
+                noisy_latents = noise_scheduler.add_noise(latents.float(), noise.float(), timesteps).to(
+                    dtype=weight_dtype
+                )
+
+                # ControlNet conditioning.
+                controlnet_image = batch["conditioning_pixel_values"].to(dtype=weight_dtype)
+                down_block_res_samples, mid_block_res_sample = controlnet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=batch["prompt_ids"],
+                    added_cond_kwargs=batch["unet_added_conditions"],
+                    controlnet_cond=controlnet_image,
+                    return_dict=False,
+                )
+
+                # Predict the noise residual
+                model_pred = unet(
+                    noisy_latents,
+                    timesteps,
+                    encoder_hidden_states=batch["prompt_ids"],
+                    added_cond_kwargs=batch["unet_added_conditions"],
+                    down_block_additional_residuals=[
+                        sample.to(dtype=weight_dtype) for sample in down_block_res_samples
+                    ],
+                    mid_block_additional_residual=mid_block_res_sample.to(dtype=weight_dtype),
+                    return_dict=False,
+                )[0]
+
+                # Get the target for loss depending on the prediction type
+                if noise_scheduler.config.prediction_type == "epsilon":
+                    target = noise
+                elif noise_scheduler.config.prediction_type == "v_prediction":
+                    target = noise_scheduler.get_velocity(latents, noise, timesteps)
+                else:
+                    raise ValueError(f"Unknown prediction type {noise_scheduler.config.prediction_type}")
+                mse_loss = F.mse_loss(model_pred.float(), target.float(), reduction="mean")
+
+
+                # print(mid_block_res_sample.shape) # torch.Size([1, 1280, 16, 16])            
+                # 分類タスクの損失を計算
+                batch_labels = []  # 正解データを格納するリスト
+                for current_prompt in batch["text"]:  # バッチ内の各プロンプトを処理
+                    label = None  # デフォルトの値を設定
+                    for class_name in class_types:
+                        if class_name in current_prompt:
+                            label = class_to_idx[class_name]  # クラス名をインデックスに変換
+                            break
+                    if label is not None:
+                        batch_labels.append(label)
+                    else:
+                        raise ValueError(f"Prompt '{current_prompt}' に該当するクラスが見つかりません。")
+                batch_labels = torch.tensor(batch_labels).to(accelerator.device)
+
+                classification_logits = classification_head(mid_block_res_sample)
+                classification_loss = F.cross_entropy(classification_logits, batch_labels)
+
+                # 主タスクと分類タスクの損失を統合
+                alpha = 0.1
+                loss = mse_loss + alpha * classification_loss  # alphaは分類タスクの重み
+
+                accelerator.backward(loss)
+                if accelerator.sync_gradients:
+                    params_to_clip = controlnet.parameters()
+                    accelerator.clip_grad_norm_(params_to_clip, args.max_grad_norm)
+                optimizer.step()
+                lr_scheduler.step()
+                optimizer.zero_grad(set_to_none=args.set_grads_to_none)
+
+            # Checks if the accelerator has performed an optimization step behind the scenes
+            if accelerator.sync_gradients:
+                progress_bar.update(1)
+                global_step += 1
+
+                # DeepSpeed requires saving weights on every device; saving weights only on the main process would cause issues.
+                if accelerator.distributed_type == DistributedType.DEEPSPEED or accelerator.is_main_process:
+                    if global_step % args.checkpointing_steps == 0:
+                        # _before_ saving state, check if this save would set us over the `checkpoints_total_limit`
+                        if args.checkpoints_total_limit is not None:
+                            checkpoints = os.listdir(args.output_dir)
+                            checkpoints = [d for d in checkpoints if d.startswith("checkpoint")]
+                            checkpoints = sorted(checkpoints, key=lambda x: int(x.split("-")[1]))
+
+                            # before we save the new checkpoint, we need to have at _most_ `checkpoints_total_limit - 1` checkpoints
+                            if len(checkpoints) >= args.checkpoints_total_limit:
+                                num_to_remove = len(checkpoints) - args.checkpoints_total_limit + 1
+                                removing_checkpoints = checkpoints[0:num_to_remove]
+
+                                logger.info(
+                                    f"{len(checkpoints)} checkpoints already exist, removing {len(removing_checkpoints)} checkpoints"
+                                )
+                                logger.info(f"removing checkpoints: {', '.join(removing_checkpoints)}")
+
+                                for removing_checkpoint in removing_checkpoints:
+                                    removing_checkpoint = os.path.join(args.output_dir, removing_checkpoint)
+                                    shutil.rmtree(removing_checkpoint)
+
+                        save_path = os.path.join(args.output_dir, f"checkpoint-{global_step}")
+                        accelerator.save_state(save_path)
+                        logger.info(f"Saved state to {save_path}")
+
+                    if args.validation_prompt is not None and global_step % args.validation_steps == 0:
+                        image_logs = log_validation(
+                            vae=vae,
+                            unet=unet,
+                            controlnet=controlnet,
+                            args=args,
+                            accelerator=accelerator,
+                            weight_dtype=weight_dtype,
+                            step=global_step,
+                        )
+
+            logs = {"mse_loss": mse_loss.detach().item(), "classification_loss": classification_loss.detach().item(), "lr": lr_scheduler.get_last_lr()[0]}
+            accelerator.log(logs, step=global_step)
+            logs["batch_labels"] = batch_labels.detach().item()
+            progress_bar.set_postfix(**logs)
+
+            if global_step >= args.max_train_steps:
+                break
+
+    # Create the pipeline using using the trained modules and save it.
+    accelerator.wait_for_everyone()
+    if accelerator.is_main_process:
+        controlnet = unwrap_model(controlnet)
+        controlnet.save_pretrained(args.output_dir)
+
+        # Run a final round of validation.
+        # Setting `vae`, `unet`, and `controlnet` to None to load automatically from `args.output_dir`.
+        image_logs = None
+        if args.validation_prompt is not None:
+            image_logs = log_validation(
+                vae=None,
+                unet=None,
+                controlnet=None,
+                args=args,
+                accelerator=accelerator,
+                weight_dtype=weight_dtype,
+                step=global_step,
+                is_final_validation=True,
+            )
+
+        if args.push_to_hub:
+            save_model_card(
+                repo_id,
+                image_logs=image_logs,
+                base_model=args.pretrained_model_name_or_path,
+                repo_folder=args.output_dir,
+            )
+            upload_folder(
+                repo_id=repo_id,
+                folder_path=args.output_dir,
+                commit_message="End of training",
+                ignore_patterns=["step_*", "epoch_*"],
+            )
+
+    accelerator.end_training()
+
+
+if __name__ == "__main__":
+    args = parse_args()
+    main(args)