\nshuffle_tags = True #@param {type:\"boolean\"}\nshuffle_caption = shuffle_tags\nactivation_tags = \"1\" #@param [0,1,2,3]\nkeep_tokens = int(activation_tags)\n\n#@markdown ### \u25b6\ufe0f Steps
\n#@markdown Your images will repeat this number of times during training. I recommend that your images multiplied by their repeats is between 200 and 400.\nnum_repeats = 10 #@param {type:\"number\"}\n#@markdown Choose how long you want to train for. A good starting point is around 10 epochs or around 2000 steps.
\n#@markdown One epoch is a number of steps equal to: your number of images multiplied by their repeats, divided by batch size.
\npreferred_unit = \"Steps\" #@param [\"Epochs\", \"Steps\"]\nhow_many = 5200 #@param {type:\"number\"}\nmax_train_epochs = how_many if preferred_unit == \"Epochs\" else None\nmax_train_steps = how_many if preferred_unit == \"Steps\" else None\n#@markdown Saving more epochs will let you compare your Lora's progress better.\nsave_every_n_epochs = 20 #@param {type:\"number\"}\nkeep_only_last_n_epochs = 5 #@param {type:\"number\"}\nif not save_every_n_epochs:\n save_every_n_epochs = max_train_epochs\nif not keep_only_last_n_epochs:\n keep_only_last_n_epochs = max_train_epochs\n\n#@markdown ### \u25b6\ufe0f Learning\n#@markdown The learning rate is the most important for your results. If your Lora produces black images, lower the unet and text encoder to 1e-4 and 1e-5 respectively, or even lower.
\n#@markdown If you're training a style you can choose to set the text encoder to 0.\nunet_lr = 3e-4 #@param {type:\"number\"}\ntext_encoder_lr = 6e-5 #@param {type:\"number\"}\n#@markdown The scheduler is the algorithm that guides the learning rate. If you're not sure, pick `constant` and ignore the number. I personally recommend `cosine_with_restarts` with 3 restarts.\nlr_scheduler = \"cosine_with_restarts\" #@param [\"constant\", \"cosine\", \"cosine_with_restarts\", \"constant_with_warmup\", \"linear\", \"polynomial\"]\nlr_scheduler_number = 3 #@param {type:\"number\"}\n#@markdown Steps spent \"warming up\" the learning rate during training for efficiency. I recommend leaving it at 5%.\nlr_warmup_ratio = 0.05 #@param {type:\"slider\", min:0.0, max:0.2, step:0.01}\nlr_warmup_steps = 0\n#@markdown Adjusts loss over time, makes learning much more efficient. The paper recommends 5.0, I recommend 8.0 for anime. Higher value makes it less strict. Set to 0 to disable.\nmin_snr_gamma = 8.0 #@param {type:\"slider\", min:0.0, max:16.0, step:0.5}\n#@markdown Multinoise may help with color balance (darker darks, lighter lights).\nmultinoise = True #@param {type:\"boolean\"}\n\n#@markdown ### \u25b6\ufe0f Structure\n#@markdown LoRA is the classic type and good for a variety of purposes. LoCon is good with artstyles as it has more layers to learn more aspects of the dataset.\nlora_type = \"LoRA\" #@param [\"LoRA\", \"LoCon\"]\n\n#@markdown Below are some recommended XL values for the following settings:\n\n#@markdown | type | network_dim | network_alpha | conv_dim | conv_alpha |\n#@markdown | :---: | :---: | :---: | :---: | :---: |\n#@markdown | Regular LoRA | 8 | 4 | | |\n#@markdown | Style LoCon | 16 | 8 | 16 | 8 |\n\n#@markdown More dim means larger Lora, it can hold more information but more isn't always better.\nnetwork_dim = 8 #@param {type:\"slider\", min:1, max:32, step:1}\nnetwork_alpha = 4 #@param {type:\"slider\", min:1, max:32, step:1}\n#@markdown The following two values only apply to the additional layers of LoCon.\nconv_dim = 4 #@param {type:\"slider\", min:1, max:32, step:1}\nconv_alpha = 1 #@param {type:\"slider\", min:1, max:32, step:1}\n\nnetwork_module = \"networks.lora\"\nnetwork_args = None\nif lora_type.lower() == \"locon\":\n network_args = [f\"conv_dim={conv_dim}\", f\"conv_alpha={conv_alpha}\"]\n\n#@markdown ### \u25b6\ufe0f Training\n#@markdown Adjust these parameters depending on your Kaggle configuration.\n#@markdown If you're using a lower tier, you should select a diffusers model at the top of this cell.\n#@markdown\n#@markdown Higher batch size is often faster but uses more memory.\ntrain_batch_size = 4 #@param {type:\"slider\", min:1, max:16, step:1}\n#@markdown I have found no substantial difference between sdpa and xformers.\ncross_attention = \"sdpa\" #@param [\"sdpa\", \"xformers\"]\n#@markdown If you're using a GPU that supports bf16, you can enable it. Otherwise, use fp16.\nmixed_precision = \"fp16\" #@param [\"bf16\", \"fp16\"]\n#@markdown Caching latents to drive will add a 250KB file next to each image but will use considerably less memory.\ncache_latents = True #@param {type:\"boolean\"}\ncache_latents_to_drive = False #@param {type:\"boolean\"} # Not applicable in Kaggle\n#@markdown The following option will turn off shuffle_tags and disable text encoder training.\ncache_text_encoder_outputs = False # @param {type:\"boolean\"}\n\n#@markdown ### \u25b6\ufe0f Advanced\n#@markdown The optimizer is the algorithm used for training. AdanW8Bit is the default and works great, while Prodigy manages learning rate automatically and may have several advantages such as training faster due to needing less steps as well as working better for small datasets.\noptimizer = \"AdamW8bit\" #@param [\"AdamW8bit\", \"Prodigy\", \"DAdaptation\", \"DadaptAdam\", \"DadaptLion\", \"AdamW\", \"Lion\", \"SGDNesterov\", \"SGDNesterov8bit\", \"AdaFactor\"]\n#@markdown Recommended args for AdamW8bit: `weight_decay=0.1 betas=[0.9,0.99]`
\n#@markdown Recommended args for Prodigy: `decouple=True weight_decay=0.01 betas=[0.9,0.999] d_coef=2 use_bias_correction=True safeguard_warmup=True`
\n#markdown Recommended args for CAME: `weight_decay=0.04`\noptimizer_args = \"weight_decay=0.1 betas=[0.9,0.99]\" #@param {type:\"string\"}\noptimizer_args = [a.strip() for a in optimizer_args.split(' ') if a]\n#@markdown If Dadapt or Prodigy are selected and the recommended box is checked, the following recommended values will override any previous settings:
\n#@markdown `unet_lr=0.75`, `text_encoder_lr=0.75`, `network_alpha=network_dim`
\n#markdown If CAME + REX is selected and the recommended box is checked, the following recommended values will override any previous settings:
\n#markdown `unet_lr=1e-4`, `text_encoder_lr=1e-6, betas=default_came_betas`\nrecommended_values = True #@param {type:\"boolean\"}\n\nif any(opt in optimizer.lower() for opt in [\"dadapt\", \"prodigy\"]):\n if recommended_values:\n unet_lr = 0.75\n text_encoder_lr = 0.75\n network_alpha = network_dim\nelif \"CAME\" in optimizer:\n optimizer = \"CAME\"\n lr_scheduler = \"REX\"\n if recommended_values:\n unet_lr = 1e-4\n text_encoder_lr = 1e-6\n for i in range(len(optimizer_args)):\n if \"betas\" in optimizer_args[i]:\n optimizer_args.pop(i)\n break\n\nlr_scheduler_num_cycles = lr_scheduler_number if lr_scheduler == \"cosine_with_restarts\" else 0\nlr_scheduler_power = lr_scheduler_number if lr_scheduler == \"polynomial\" else 0\n\n#@markdown ### \u25b6\ufe0f Ready\n#@markdown You can now run this cell to cook your Lora. Good luck!
\n\n\n# \ud83d\udc69\u200d\ud83d\udcbb Cool code goes here\n\n# Define root directory for Kaggle\nroot_dir = \"/kaggle/working\"\ndeps_dir = os.path.join(root_dir, \"deps\")\nrepo_dir = os.path.join(root_dir, \"kohya-trainer\")\n\nif \"/Loras\" in folder_structure:\n main_dir = os.path.join(root_dir, \"Loras\")\n log_folder = os.path.join(main_dir, \"_logs\")\n config_folder = os.path.join(main_dir, project_name)\n images_folder = os.path.join(main_dir, project_name, \"dataset\")\n output_folder = os.path.join(main_dir, project_name, \"output\")\nelse:\n main_dir = os.path.join(root_dir, \"lora_training\")\n images_folder = os.path.join(main_dir, \"datasets\", project_name)\n output_folder = os.path.join(main_dir, \"output\", project_name)\n config_folder = os.path.join(main_dir, \"config\", project_name)\n log_folder = os.path.join(main_dir, \"log\")\n\nconfig_file = os.path.join(config_folder, \"training_config.toml\")\ndataset_config_file = os.path.join(config_folder, \"dataset_config.toml\")\naccelerate_config_file = os.path.join(repo_dir, \"accelerate_config/config.yaml\")\n\ndef install_dependencies():\n os.chdir(root_dir)\n if not os.path.exists(repo_dir):\n !git clone {SOURCE} {repo_dir}\n os.chdir(repo_dir)\n if COMMIT:\n !git reset --hard {COMMIT}\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/train_network_xl_wrapper.py -q -O train_network_xl_wrapper.py\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/dracula.py -q -O dracula.py\n\n # Kaggle may restrict apt installations; proceed if necessary\n # !apt -y update -qq\n # !apt -y install aria2 -qq\n\n # Install necessary Python packages\n !pip install --quiet torch==2.5.1+cu121 accelerate==0.19.0 transformers==4.30.2 diffusers==0.18.2 \\\n bitsandbytes==0.40.0.post4 opencv-python jax==0.4.23 jaxlib==0.4.23 \\\n pytorch-lightning==1.9.0 voluptuous==0.13.1 toml==0.10.2 ftfy==6.1.1 einops==0.6.0 \\\n safetensors pygments huggingface-hub==0.22.0 wandb invisible-watermark==0.2.0 open-clip-torch==2.20.0 \\\n dadaptation==3.1 prodigyopt==1.0 lion-pytorch==0.1.2\n !pip install -e .\n if cross_attention == \"xformers\":\n !pip install -q xformers==0.0.26.dev778\n if \"CAME\" in optimizer:\n !pip install came-pytorch\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/train_util.py -q -O library/train_util.py\n\n # Patch kohya for minor stuff\n if LOWRAM:\n !sed -i \"s@cpu@cuda@\" library/model_util.py\n if LOAD_TRUNCATED_IMAGES:\n !sed -i 's/from PIL import Image/from PIL import Image, ImageFile\\nImageFile.LOAD_TRUNCATED_IMAGES=True/g' library/train_util.py # fix truncated jpegs error\n if BETTER_EPOCH_NAMES:\n !sed -i 's/{:06d}/{:02d}/g' library/train_util.py # make epoch names shorter\n !sed -i 's/\".\" + args.save_model_as)/\"-{:02d}.\".format(num_train_epochs) + args.save_model_as)/g' train_network.py # name of the last epoch will match the rest\n\n from accelerate.utils import write_basic_config\n if not os.path.exists(accelerate_config_file):\n write_basic_config(save_location=accelerate_config_file)\n\n os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"3\"\n os.environ[\"BITSANDBYTES_NOWELCOME\"] = \"1\"\n os.environ[\"SAFETENSORS_FAST_GPU\"] = \"1\"\n\ndef validate_dataset():\n global lr_warmup_steps, lr_warmup_ratio, caption_extension, keep_tokens, model_url\n supported_types = (\".png\", \".jpg\", \".jpeg\", \".webp\", \".bmp\")\n\n if model_url.startswith(\"/kaggle/input/\") and not os.path.exists(model_url):\n print(\"\ud83d\udca5 Error: The custom training model you specified was not found in your Kaggle input.\")\n return\n\n print(\"\\n\ud83d\udcbf Checking dataset...\")\n if not project_name.strip() or any(c in project_name for c in \" .()\\\"'\\\\/\"):\n print(\"\ud83d\udca5 Error: Please choose a valid project name.\")\n return\n\n # Find the folders and files\n if custom_dataset:\n try:\n datconf = toml.loads(custom_dataset)\n datasets = [d for d in datconf[\"datasets\"][0][\"subsets\"]]\n except:\n print(f\"\ud83d\udca5 Error: Your custom dataset is invalid or contains an error! Please check the original template.\")\n return\n reg = [d.get(\"image_dir\") for d in datasets if d.get(\"is_reg\", False)]\n datasets_dict = {d[\"image_dir\"]: d[\"num_repeats\"] for d in datasets}\n folders = datasets_dict.keys()\n files = [f for folder in folders for f in os.listdir(folder)]\n images_repeats = {folder: (len([f for f in os.listdir(folder) if f.lower().endswith(supported_types)]), datasets_dict[folder]) for folder in folders}\n else:\n reg = []\n folders = [images_folder]\n if not os.path.exists(images_folder):\n print(f\"\ud83d\udca5 Error: Images folder {images_folder} does not exist.\")\n return\n files = os.listdir(images_folder)\n images_repeats = {images_folder: (len([f for f in files if f.lower().endswith(supported_types)]), num_repeats)}\n\n # Validation\n for folder in folders:\n if not os.path.exists(folder):\n print(f\"\ud83d\udca5 Error: The folder {folder} doesn't exist.\")\n return\n for folder, (img, rep) in images_repeats.items():\n if not img:\n print(f\"\ud83d\udca5 Error: Your {folder} folder is empty.\")\n return\n test_files = []\n for f in files:\n if not f.lower().endswith((caption_extension, \".npz\")) and not f.lower().endswith(supported_types):\n print(f\"\ud83d\udca5 Error: Invalid file in dataset: \\\"{f}\\\". Aborting.\")\n return\n for ff in test_files:\n if f.endswith(supported_types) and ff.endswith(supported_types) \\\n and os.path.splitext(f)[0] == os.path.splitext(ff)[0]:\n print(f\"\ud83d\udca5 Error: The files {f} and {ff} cannot have the same name. Aborting.\")\n return\n test_files.append(f)\n\n if caption_extension and not [txt for txt in files if txt.lower().endswith(caption_extension)]:\n caption_extension = \"\"\n if continue_from_lora and not (continue_from_lora.endswith(\".safetensors\") and os.path.exists(continue_from_lora)):\n print(f\"\ud83d\udca5 Error: Invalid path to existing Lora. Example: /kaggle/working/Loras/example.safetensors\")\n return\n\n # Pretty stuff\n\n pre_steps_per_epoch = sum(img * rep for (img, rep) in images_repeats.values())\n steps_per_epoch = pre_steps_per_epoch / train_batch_size\n total_steps = max_train_steps or int(max_train_epochs * steps_per_epoch)\n estimated_epochs = int(total_steps / steps_per_epoch)\n lr_warmup_steps = int(total_steps * lr_warmup_ratio)\n\n for folder, (img, rep) in images_repeats.items():\n print(\"\ud83d\udcc1 \" + folder.replace(\"/kaggle/input/\", \"\") + (\" (Regularization)\" if folder in reg else \"\"))\n print(f\"\ud83d\udcc8 Found {img} images with {rep} repeats, equaling {img * rep} steps.\")\n print(f\"\ud83d\udcc9 Divide {pre_steps_per_epoch} steps by {train_batch_size} batch size to get {steps_per_epoch} steps per epoch.\")\n if max_train_epochs:\n print(f\"\ud83d\udd2e There will be {max_train_epochs} epochs, for around {total_steps} total training steps.\")\n else:\n print(f\"\ud83d\udd2e There will be {total_steps} steps, divided into {estimated_epochs} epochs and then some.\")\n\n if total_steps > 10000:\n print(\"\ud83d\udca5 Error: Your total steps are too high. You probably made a mistake. Aborting...\")\n return\n\n return True\n\ndef create_config():\n global dataset_config_file, config_file, model_file\n\n if override_config_file:\n config_file = override_config_file\n print(f\"\\n\u2b55 Using custom config file {config_file}\")\n else:\n config_dict = {\n \"network_arguments\": {\n \"unet_lr\": unet_lr,\n \"text_encoder_lr\": text_encoder_lr if not cache_text_encoder_outputs else 0,\n \"network_dim\": network_dim,\n \"network_alpha\": network_alpha,\n \"network_module\": network_module,\n \"network_args\": network_args,\n \"network_train_unet_only\": text_encoder_lr == 0 or cache_text_encoder_outputs,\n \"network_weights\": continue_from_lora if continue_from_lora else None\n },\n \"optimizer_arguments\": {\n \"learning_rate\": unet_lr,\n \"lr_scheduler\": lr_scheduler,\n \"lr_scheduler_num_cycles\": lr_scheduler_num_cycles if lr_scheduler == \"cosine_with_restarts\" else None,\n \"lr_scheduler_power\": lr_scheduler_power if lr_scheduler == \"polynomial\" else None,\n \"lr_warmup_steps\": lr_warmup_steps if lr_scheduler != \"constant\" else None,\n \"optimizer_type\": optimizer,\n \"optimizer_args\": optimizer_args if optimizer_args else None,\n },\n \"training_arguments\": {\n \"pretrained_model_name_or_path\": model_file,\n \"vae\": vae_file,\n \"max_train_steps\": max_train_steps,\n \"max_train_epochs\": max_train_epochs,\n \"train_batch_size\": train_batch_size,\n \"seed\": 42,\n \"max_token_length\": 225,\n \"xformers\": cross_attention == \"xformers\",\n \"sdpa\": cross_attention == \"sdpa\",\n \"min_snr_gamma\": min_snr_gamma if min_snr_gamma > 0 else None,\n \"lowram\": LOWRAM,\n \"no_half_vae\": True,\n \"gradient_checkpointing\": True,\n \"gradient_accumulation_steps\": 1,\n \"max_data_loader_n_workers\": 8,\n \"persistent_data_loader_workers\": True,\n \"mixed_precision\": mixed_precision,\n \"full_bf16\": mixed_precision == \"bf16\",\n \"cache_latents\": cache_latents,\n \"cache_latents_to_disk\": cache_latents_to_drive,\n \"cache_text_encoder_outputs\": cache_text_encoder_outputs,\n \"min_timestep\": 0,\n \"max_timestep\": 1000,\n \"prior_loss_weight\": 1.0,\n \"multires_noise_iterations\": 6 if multinoise else None,\n \"multires_noise_discount\": 0.3 if multinoise else None,\n },\n \"saving_arguments\": {\n \"save_precision\": \"fp16\",\n \"save_model_as\": \"safetensors\",\n \"save_every_n_epochs\": save_every_n_epochs,\n \"save_last_n_epochs\": keep_only_last_n_epochs,\n \"output_name\": project_name,\n \"output_dir\": output_folder,\n \"log_prefix\": project_name,\n \"logging_dir\": log_folder,\n \"wandb_api_key\": wandb_key if wandb_key else None,\n \"log_with\": \"wandb\" if wandb_key else None,\n }\n }\n\n for key in config_dict:\n if isinstance(config_dict[key], dict):\n config_dict[key] = {k: v for k, v in config_dict[key].items() if v is not None}\n\n with open(config_file, \"w\") as f:\n f.write(toml.dumps(config_dict))\n print(f\"\\n\ud83d\udcc4 Config saved to {config_file}\")\n\n if override_dataset_config_file:\n dataset_config_file = override_dataset_config_file\n print(f\"\u2b55 Using custom dataset config file {dataset_config_file}\")\n else:\n dataset_config_dict = {\n \"general\": {\n \"resolution\": resolution,\n \"shuffle_caption\": shuffle_caption and not cache_text_encoder_outputs,\n \"keep_tokens\": keep_tokens,\n \"flip_aug\": False,\n \"caption_extension\": caption_extension,\n \"enable_bucket\": True,\n \"bucket_no_upscale\": False,\n \"bucket_reso_steps\": 64,\n \"min_bucket_reso\": 256,\n \"max_bucket_reso\": 4096,\n },\n \"datasets\": toml.loads(custom_dataset)[\"datasets\"] if custom_dataset else [\n {\n \"subsets\": [\n {\n \"num_repeats\": num_repeats,\n \"image_dir\": images_folder,\n \"class_tokens\": None if caption_extension else project_name\n }\n ]\n }\n ]\n }\n\n for key in dataset_config_dict:\n if isinstance(dataset_config_dict[key], dict):\n dataset_config_dict[key] = {k: v for k, v in dataset_config_dict[key].items() if v is not None}\n\n with open(dataset_config_file, \"w\") as f:\n f.write(toml.dumps(dataset_config_dict))\n print(f\"\ud83d\udcc4 Dataset config saved to {dataset_config_file}\")\n\ndef download_model():\n global old_model_url, model_url, model_file, vae_url, vae_file\n real_model_url = model_url # There was a reason for having a separate variable but I forgot what it was.\n\n if real_model_url.startswith(\"/kaggle/input/\") and not os.path.exists(real_model_url):\n print(\"\ud83d\udca5 Error: The custom training model you specified was not found in your Kaggle input.\")\n return\n\n if not load_diffusers:\n if real_model_url.startswith(\"/kaggle/input/\"):\n # Local model, already checked to exist\n model_file = real_model_url\n print(f\"\ud83d\udcc1 Using local model file: {model_file}\")\n # Validation\n if model_file.lower().endswith(\".safetensors\"):\n from safetensors.torch import load_file as load_safetensors\n try:\n test = load_safetensors(model_file)\n del test\n except:\n return False\n elif model_file.lower().endswith(\".ckpt\"):\n from torch import load as load_ckpt\n try:\n test = load_ckpt(model_file)\n del test\n except:\n return False\n return True\n\n # Downloadable model\n if load_diffusers:\n if 'huggingface.co' in real_model_url:\n match = re.search(r'huggingface\\.co/([^/]+)/([^/]+)', real_model_url)\n if match:\n username = match.group(1)\n model_name = match.group(2)\n model_file = f\"{username}/{model_name}\"\n fs = HfFileSystem()\n existing_folders = set(fs.ls(model_file, detail=False))\n necessary_folders = [\"scheduler\", \"text_encoder\", \"text_encoder_2\", \"tokenizer\", \"tokenizer_2\", \"unet\", \"vae\"]\n if all(f\"{model_file}/{folder}\" in existing_folders for folder in necessary_folders):\n print(\"\ud83c\udf43 Diffusers model identified.\") # Will be handled by kohya\n return True\n raise ValueError(\"\ud83d\udca5 Failed to load Diffusers model. If this model is not Diffusers, have you tried turning it off at the top of this cell?\")\n\n # Define local filename\n if not model_file:\n if real_model_url.lower().endswith((\".ckpt\", \".safetensors\")):\n model_file = os.path.join(root_dir, os.path.basename(real_model_url))\n else:\n model_file = os.path.join(root_dir, \"downloaded_model.safetensors\")\n if os.path.exists(model_file):\n os.remove(model_file)\n\n # HuggingFace\n if m := re.search(r\"(?:https?://)?(?:www\\.)?huggingface\\.co/[^/]+/[^/]+/blob\", real_model_url):\n real_model_url = real_model_url.replace(\"blob\", \"resolve\")\n # Civitai\n elif m := re.search(r\"(?:https?://)?(?:www\\.)?civitai\\.com/models/([0-9]+)(/[A-Za-z0-9-_]+)?\", real_model_url):\n if m.group(2):\n model_file = os.path.join(root_dir, f\"{m.group(2)}.safetensors\")\n if m := re.search(r\"modelVersionId=([0-9]+)\", real_model_url):\n real_model_url = f\"https://civitai.com/api/download/models/{m.group(1)}\"\n else:\n raise ValueError(\"\ud83d\udca5 optional_custom_training_model contains a civitai link, but the link doesn't include a modelVersionId. You can also right click the download button to copy the direct download link.\")\n\n # Download checkpoint\n !wget \"{real_model_url}\" -O \"{model_file}\" --quiet\n\n # Download VAE\n if not os.path.exists(vae_file):\n !wget \"{vae_url}\" -O \"{vae_file}\" --quiet\n\n # Validation\n\n if model_file.lower().endswith(\".safetensors\"):\n from safetensors.torch import load_file as load_safetensors\n try:\n test = load_safetensors(model_file)\n del test\n except:\n new_model_file = os.path.splitext(model_file)[0] + \".ckpt\"\n os.rename(model_file, new_model_file)\n model_file = new_model_file\n print(f\"Renamed model to {os.path.splitext(model_file)[0]}.ckpt\")\n\n if model_file.lower().endswith(\".ckpt\"):\n from torch import load as load_ckpt\n try:\n test = load_ckpt(model_file)\n del test\n except:\n return False\n\n return True\n\ndef main():\n global dependencies_installed\n\n # Create all necessary directories\n for dir in (main_dir, deps_dir, repo_dir, log_folder, images_folder, output_folder, config_folder):\n os.makedirs(dir, exist_ok=True)\n\n if not validate_dataset():\n return\n\n if not dependencies_installed:\n print(\"\\n\ud83c\udfed Installing dependencies...\\n\")\n t0 = time()\n install_dependencies()\n t1 = time()\n dependencies_installed = True\n print(f\"\\n\u2705 Installation finished in {int(t1 - t0)} seconds.\")\n else:\n print(\"\\n\u2705 Dependencies already installed.\")\n\n if old_model_url != model_url or not model_file or not os.path.exists(model_file):\n print(\"\\n\ud83d\udd04 Getting model...\")\n if not download_model():\n print(\"\\n\ud83d\udca5 Error: The model you specified is invalid or corrupted.\"\n \"\\nIf you're using an URL, please check that the model is accessible without being logged in.\"\n \"\\nYou can try civitai or huggingface URLs, or a path in your Kaggle input starting with /kaggle/input\")\n return\n print()\n else:\n print(\"\\n\ud83d\udd04 Model already downloaded.\\n\")\n\n create_config()\n\n print(\"\\n\u2b50 Starting trainer...\\n\")\n os.chdir(repo_dir)\n\n # Execute the training script\n !accelerate launch --quiet --config_file={accelerate_config_file} --num_cpu_threads_per_process=1 train_network_xl_wrapper.py --dataset_config={dataset_config_file} --config_file={config_file}\n\n # Check for successful completion\n exit_code = os.WEXITSTATUS(os.system('echo $?'))\n if exit_code == 0:\n display(Markdown(f\"### \u2705 Done! [Download your Lora from /kaggle/working/{output_folder}]\\n\"\n \"### There will be several files, you should try the latest version (the file with the largest number next to it)\"))\n else:\n print(\"\ud83d\udca5 Error during training. Please check the logs for more details.\")\n\nmain()\n```\n\n### **Additional Notes:**\n\n1. **Data Upload:**\n - **Kaggle Datasets:** Upload your dataset and models as Kaggle Datasets and reference them using `/kaggle/input/`. Alternatively, you can upload them directly to the notebook environment.\n \n2. **Storage Limitations:**\n - **Kaggle's Storage:** Ensure that your data and models fit within Kaggle's storage limits. Kaggle typically provides up to 20GB of storage per notebook.\n\n3. **Runtime Limits:**\n - **Session Duration:** Kaggle notebooks have a maximum runtime. Ensure your training completes within the allowed time or consider checkpointing your progress.\n\n4. **GPU Availability:**\n - **Hardware Selection:** When setting up your Kaggle notebook, choose the appropriate GPU (e.g., Nvidia Tesla P100) based on your training requirements.\n\n5. **Environment Variables:**\n - **Hugging Face Authentication:** If your model requires authentication, ensure you set up the necessary environment variables or use authentication tokens appropriately.\n\n6. **Debugging:**\n - **Logs:** Use Kaggle's output logs to debug any issues during training. Adjust configurations as needed based on the feedback.\n\nBy implementing these changes, you should be able to run the training script seamlessly on Kaggle. Ensure that all dependencies are compatible and that paths correctly reference Kaggle's directory structure.","behavior":"degenerate_repetition","conversation_hash":"a18138fc6c1f50ed14516ae5ca2ee654","conversation_redacted":false,"conversation_toxic":false,"country":"Pakistan","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"o1-mini-2024-09-12","n_chars":60491,"n_turns_in_conversation":1,"n_words":6118,"source_row_offset":28607,"source_shard":"turns-00052.parquet","timestamp":"Sat, 16 Nov 2024 10:34:30 GMT","turn_id":"f1927e2a0ce1e606ead27ac0","turn_index":0,"user_text":"edit this colab code to run on kaggle, import os\nimport re\nimport toml\nimport pathlib\nfrom time import time\nfrom huggingface_hub import HfFileSystem\nfrom IPython.display import Markdown, display\n\n# These carry information from past executions\nif \"model_url\" in globals():\n old_model_url = model_url\nelse:\n old_model_url = None\nif \"dependencies_installed\" not in globals():\n dependencies_installed = False\nif \"model_file\" not in globals():\n model_file = None\n\n# These may be set by other cells, some are legacy\nif \"custom_dataset\" not in globals():\n custom_dataset = None\nif \"override_dataset_config_file\" not in globals():\n override_dataset_config_file = None\nif \"continue_from_lora\" not in globals():\n continue_from_lora = \"\"\nif \"override_config_file\" not in globals():\n override_config_file = None\n\nCOLAB = True\nSOURCE = \"https://github.com/qaneel/kohya-trainer\"\nCOMMIT = None\nBETTER_EPOCH_NAMES = True\nLOAD_TRUNCATED_IMAGES = True\ntry:\n LOWRAM = int(next(line.split()[1] for line in open('/proc/meminfo') if \"MemTotal\" in line)) / (1024**2) < 15\nexcept:\n LOWRAM = False\n\n#@title ## \ud83d\udea9 Start Here\n\n#@markdown ### \u25b6\ufe0f Setup\n#@markdown Your project name will be the same as the folder containing your images. Spaces aren't allowed.\nproject_name = \"S1s\" #@param {type:\"string\"}\nproject_name = project_name.strip()\n#@markdown The folder structure doesn't matter and is purely for comfort. Make sure to always pick the same one. I like organizing by project.\nfolder_structure = \"Organize by project (MyDrive/Loras/project_name/dataset)\" #@param [\"Organize by category (MyDrive/lora_training/datasets/project_name)\", \"Organize by project (MyDrive/Loras/project_name/dataset)\"]\n#@markdown Decide the model that will be downloaded and used for training. You can also choose your own by pasting its download link, or a file in your Google Drive starting with `/content/drive/MyDrive`.\ntraining_model = \"Stable Diffusion XL 1.0 base\" #@param [\"Pony Diffusion V6 XL\", \"Illustrious XL 0.1\", \"Animagine XL V3\", \"Stable Diffusion XL 1.0 base\"]\noptional_custom_training_model = \"\" #@param {type:\"string\"}\n#@markdown Using a diffusers model utilizes less resources. The 3 options above will work with or without diffusers.\nload_diffusers = True #@param {type:\"boolean\"}\n#@markdown Use wandb if you want to visualize the progress of your training over time.\nwandb_key = \"\" #@param {type:\"string\"}\n\nif optional_custom_training_model:\n model_url = optional_custom_training_model\nelif \"Pony\" in training_model:\n if load_diffusers:\n model_url = \"https://huggingface.co/hollowstrawberry/67AB2F\"\n else:\n model_url = \"https://civitai.com/api/download/models/290640\"\n model_file = \"/content/ponyDiffusionV6XL.safetensors\"\nelif \"Animagine\" in training_model:\n if load_diffusers:\n model_url = \"https://huggingface.co/cagliostrolab/animagine-xl-3.0\"\n else:\n model_url = \"https://civitai.com/api/download/models/293564\"\n model_file = \"/content/animagineXLV3.safetensors\"\nelif \"Illustrious\" in training_model:\n if load_diffusers:\n model_url = \"https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0\"\n else:\n model_url = \"https://huggingface.co/OnomaAIResearch/Illustrious-xl-early-release-v0/resolve/main/Illustrious-XL-v0.1.safetensors\"\nelse:\n if load_diffusers:\n model_url = \"https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/\"\n else:\n model_url = \"https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/resolve/main/sd_xl_base_1.0.safetensors\"\n\nif load_diffusers:\n vae_file= \"stabilityai/sdxl-vae\"\nelse:\n vae_url = \"https://huggingface.co/stabilityai/sdxl-vae/resolve/main/sdxl_vae.safetensors\"\n vae_file = \"/content/sdxl_vae.safetensors\"\n\nmodel_url = model_url.strip()\n\n#@markdown ### \u25b6\ufe0f Processing\nresolution = 1024 #param {type:\"slider\", min:768, max:1536, step:128}\ncaption_extension = \".txt\" #@param [\".txt\", \".caption\"]\n#@markdown Shuffling anime tags in place improves learning and prompting. An activation tag goes at the start of every text file and will not be shuffled.
\nshuffle_tags = True #@param {type:\"boolean\"}\nshuffle_caption = shuffle_tags\nactivation_tags = \"1\" #@param [0,1,2,3]\nkeep_tokens = int(activation_tags)\n\n#@markdown ### \u25b6\ufe0f Steps
\n#@markdown Your images will repeat this number of times during training. I recommend that your images multiplied by their repeats is between 200 and 400.\nnum_repeats = 10 #@param {type:\"number\"}\n#@markdown Choose how long you want to train for. A good starting point is around 10 epochs or around 2000 steps.
\n#@markdown One epoch is a number of steps equal to: your number of images multiplied by their repeats, divided by batch size.
\npreferred_unit = \"Steps\" #@param [\"Epochs\", \"Steps\"]\nhow_many = 5200 #@param {type:\"number\"}\nmax_train_epochs = how_many if preferred_unit == \"Epochs\" else None\nmax_train_steps = how_many if preferred_unit == \"Steps\" else None\n#@markdown Saving more epochs will let you compare your Lora's progress better.\nsave_every_n_epochs = 20 #@param {type:\"number\"}\nkeep_only_last_n_epochs = 5 #@param {type:\"number\"}\nif not save_every_n_epochs:\n save_every_n_epochs = max_train_epochs\nif not keep_only_last_n_epochs:\n keep_only_last_n_epochs = max_train_epochs\n\n#@markdown ### \u25b6\ufe0f Learning\n#@markdown The learning rate is the most important for your results. If your Lora produces black images, lower the unet and text encoder to 1e-4 and 1e-5 respectively, or even lower.
\n#@markdown If you're training a style you can choose to set the text encoder to 0.\nunet_lr = 3e-4 #@param {type:\"number\"}\ntext_encoder_lr = 6e-5 #@param {type:\"number\"}\n#@markdown The scheduler is the algorithm that guides the learning rate. If you're not sure, pick `constant` and ignore the number. I personally recommend `cosine_with_restarts` with 3 restarts.\nlr_scheduler = \"cosine_with_restarts\" #@param [\"constant\", \"cosine\", \"cosine_with_restarts\", \"constant_with_warmup\", \"linear\", \"polynomial\"]\nlr_scheduler_number = 3 #@param {type:\"number\"}\n#@markdown Steps spent \"warming up\" the learning rate during training for efficiency. I recommend leaving it at 5%.\nlr_warmup_ratio = 0.05 #@param {type:\"slider\", min:0.0, max:0.2, step:0.01}\nlr_warmup_steps = 0\n#@markdown Adjusts loss over time, makes learning much more efficient. The paper recommends 5.0, I recommend 8.0 for anime. Higher value makes it less strict. Set to 0 to disable.\nmin_snr_gamma = 8.0 #@param {type:\"slider\", min:0.0, max:16.0, step:0.5}\n#@markdown Multinoise may help with color balance (darker darks, lighter lights).\nmultinoise = True #@param {type:\"boolean\"}\n\n#@markdown ### \u25b6\ufe0f Structure\n#@markdown LoRA is the classic type and good for a variety of purposes. LoCon is good with artstyles as it has more layers to learn more aspects of the dataset.\nlora_type = \"LoRA\" #@param [\"LoRA\", \"LoCon\"]\n\n#@markdown Below are some recommended XL values for the following settings:\n\n#@markdown | type | network_dim | network_alpha | conv_dim | conv_alpha |\n#@markdown | :---: | :---: | :---: | :---: | :---: |\n#@markdown | Regular LoRA | 8 | 4 | | |\n#@markdown | Style LoCon | 16 | 8 | 16 | 8 |\n\n#@markdown More dim means larger Lora, it can hold more information but more isn't always better.\nnetwork_dim = 8 #@param {type:\"slider\", min:1, max:32, step:1}\nnetwork_alpha = 4 #@param {type:\"slider\", min:1, max:32, step:1}\n#@markdown The following two values only apply to the additional layers of LoCon.\nconv_dim = 4 #@param {type:\"slider\", min:1, max:32, step:1}\nconv_alpha = 1 #@param {type:\"slider\", min:1, max:32, step:1}\n\nnetwork_module = \"networks.lora\"\nnetwork_args = None\nif lora_type.lower() == \"locon\":\n network_args = [f\"conv_dim={conv_dim}\", f\"conv_alpha={conv_alpha}\"]\n\n#@markdown ### \u25b6\ufe0f Training\n#@markdown Adjust these parameters depending on your colab configuration.\n#@markdown If you're using the free tier you should select a diffusers model at the top of this cell.\n#@markdown\n#@markdown Higher batch size is often faster but uses more memory.\ntrain_batch_size = 4 #@param {type:\"slider\", min:1, max:16, step:1}\n#@markdown I have found no substantial difference between sdpa and xformers.\ncross_attention = \"sdpa\" #@param [\"sdpa\", \"xformers\"]\n#@markdown If you're on an A100 you should enable bf16.\nmixed_precision = \"fp16\" #@param [\"bf16\", \"fp16\"]\n#@markdown Caching latents to drive will add a 250KB file next to each image but will use considerably less memory.\ncache_latents = True #@param {type:\"boolean\"}\ncache_latents_to_drive = True #@param {type:\"boolean\"}\n#@markdown The following option will turn off shuffle_tags and disable text encoder training.\ncache_text_encoder_outputs = False # @param {type:\"boolean\"}\n\n#@markdown ### \u25b6\ufe0f Advanced\n#@markdown The optimizer is the algorithm used for training. AdanW8Bit is the default and works great, while Prodigy manages learning rate automatically and may have several advantages such as training faster due to needing less steps as well as working better for small datasets.\noptimizer = \"AdamW8bit\" #@param [\"AdamW8bit\", \"Prodigy\", \"DAdaptation\", \"DadaptAdam\", \"DadaptLion\", \"AdamW\", \"Lion\", \"SGDNesterov\", \"SGDNesterov8bit\", \"AdaFactor\"]\n#@markdown Recommended args for AdamW8bit: `weight_decay=0.1 betas=[0.9,0.99]`
\n#@markdown Recommended args for Prodigy: `decouple=True weight_decay=0.01 betas=[0.9,0.999] d_coef=2 use_bias_correction=True safeguard_warmup=True`
\n#markdown Recommended args for CAME: `weight_decay=0.04`\noptimizer_args = \"weight_decay=0.1 betas=[0.9,0.99]\" #@param {type:\"string\"}\noptimizer_args = [a.strip() for a in optimizer_args.split(' ') if a]\n#@markdown If Dadapt or Prodigy are selected and the recommended box is checked, the following recommended values will override any previous settings:
\n#@markdown `unet_lr=0.75`, `text_encoder_lr=0.75`, `network_alpha=network_dim`
\n#markdown If CAME + REX is selected and the recommended box is checked, the following recommended values will override any previous settings:
\n#markdown `unet_lr=1e-4`, `text_encoder_lr=1e-6, betas=default_came_betas`\nrecommended_values = True #@param {type:\"boolean\"}\n\nif any(opt in optimizer.lower() for opt in [\"dadapt\", \"prodigy\"]):\n if recommended_values:\n unet_lr = 0.75\n text_encoder_lr = 0.75\n network_alpha = network_dim\nelif \"CAME\" in optimizer:\n optimizer = \"CAME\"\n lr_scheduler = \"REX\"\n if recommended_values:\n unet_lr = 1e-4\n text_encoder_lr = 1e-6\n for i in range(len(optimizer_args)):\n if \"betas\" in optimizer_args[i]:\n optimizer_args.pop(i)\n break\n\nlr_scheduler_num_cycles = lr_scheduler_number if lr_scheduler == \"cosine_with_restarts\" else 0\nlr_scheduler_power = lr_scheduler_number if lr_scheduler == \"polynomial\" else 0\n\n\n#@markdown ### \u25b6\ufe0f Ready\n#@markdown You can now run this cell to cook your Lora. Good luck!
\n\n\n# \ud83d\udc69\u200d\ud83d\udcbb Cool code goes here\n\nroot_dir = \"/content\" if COLAB else pathlib.Path.home() / \"Loras\"\ndeps_dir = os.path.join(root_dir, \"deps\")\nrepo_dir = os.path.join(root_dir, \"kohya-trainer\")\n\nif \"/Loras\" in folder_structure:\n main_dir = os.path.join(root_dir, \"drive/MyDrive/Loras\") if COLAB else root_dir\n log_folder = os.path.join(main_dir, \"_logs\")\n config_folder = os.path.join(main_dir, project_name)\n images_folder = os.path.join(main_dir, project_name, \"dataset\")\n output_folder = os.path.join(main_dir, project_name, \"output\")\nelse:\n main_dir = os.path.join(root_dir, \"drive/MyDrive/lora_training\") if COLAB else root_dir\n images_folder = os.path.join(main_dir, \"datasets\", project_name)\n output_folder = os.path.join(main_dir, \"output\", project_name)\n config_folder = os.path.join(main_dir, \"config\", project_name)\n log_folder = os.path.join(main_dir, \"log\")\n\nconfig_file = os.path.join(config_folder, \"training_config.toml\")\ndataset_config_file = os.path.join(config_folder, \"dataset_config.toml\")\naccelerate_config_file = os.path.join(repo_dir, \"accelerate_config/config.yaml\")\n\ndef install_dependencies():\n os.chdir(root_dir)\n !git clone {SOURCE} {repo_dir}\n os.chdir(repo_dir)\n if COMMIT:\n !git reset --hard {COMMIT}\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/train_network_xl_wrapper.py -q -O train_network_xl_wrapper.py\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/dracula.py -q -O dracula.py\n\n !apt -y update -qq\n !apt -y install aria2 -qq\n !pip install torch==2.5.1+cu121 accelerate==0.19.0 transformers==4.30.2 diffusers==0.18.2 \\\n bitsandbytes==0.40.0.post4 opencv-python jax==0.4.23 jaxlib==0.4.23 \\\n pytorch-lightning==1.9.0 voluptuous==0.13.1 toml==0.10.2 ftfy==6.1.1 einops==0.6.0 \\\n safetensors pygments huggingface-hub==0.22.0 wandb invisible-watermark==0.2.0 open-clip-torch==2.20.0 \\\n dadaptation==3.1 prodigyopt==1.0 lion-pytorch==0.1.2\n !pip install -e .\n if cross_attention == \"xformers\":\n !pip install -q xformers==0.0.26.dev778\n if \"CAME\" in optimizer:\n !pip install came-pytorch\n !wget https://raw.githubusercontent.com/hollowstrawberry/kohya-colab/main/train_util.py -q -O library/train_util.py\n\n # patch kohya for minor stuff\n if LOWRAM:\n !sed -i \"s@cpu@cuda@\" library/model_util.py\n if LOAD_TRUNCATED_IMAGES:\n !sed -i 's/from PIL import Image/from PIL import Image, ImageFile\\nImageFile.LOAD_TRUNCATED_IMAGES=True/g' library/train_util.py # fix truncated jpegs error\n if BETTER_EPOCH_NAMES:\n !sed -i 's/{:06d}/{:02d}/g' library/train_util.py # make epoch names shorter\n !sed -i 's/\".\" + args.save_model_as)/\"-{:02d}.\".format(num_train_epochs) + args.save_model_as)/g' train_network.py # name of the last epoch will match the rest\n\n from accelerate.utils import write_basic_config\n if not os.path.exists(accelerate_config_file):\n write_basic_config(save_location=accelerate_config_file)\n\n os.environ[\"TF_CPP_MIN_LOG_LEVEL\"] = \"3\"\n os.environ[\"BITSANDBYTES_NOWELCOME\"] = \"1\"\n os.environ[\"SAFETENSORS_FAST_GPU\"] = \"1\"\n\ndef validate_dataset():\n global lr_warmup_steps, lr_warmup_ratio, caption_extension, keep_tokens, model_url\n supported_types = (\".png\", \".jpg\", \".jpeg\", \".webp\", \".bmp\")\n\n if model_url.startswith(\"/content/drive/\") and not os.path.exists(model_url):\n print(\"\ud83d\udca5 Error: The custom training model you specified was not found in your Google Drive.\")\n return\n\n print(\"\\n\ud83d\udcbf Checking dataset...\")\n if not project_name.strip() or any(c in project_name for c in \" .()\\\"'\\\\/\"):\n print(\"\ud83d\udca5 Error: Please choose a valid project name.\")\n return\n\n # Find the folders and files\n if custom_dataset:\n try:\n datconf = toml.loads(custom_dataset)\n datasets = [d for d in datconf[\"datasets\"][0][\"subsets\"]]\n except:\n print(f\"\ud83d\udca5 Error: Your custom dataset is invalid or contains an error! Please check the original template.\")\n return\n reg = [d.get(\"image_dir\") for d in datasets if d.get(\"is_reg\", False)]\n datasets_dict = {d[\"image_dir\"]: d[\"num_repeats\"] for d in datasets}\n folders = datasets_dict.keys()\n files = [f for folder in folders for f in os.listdir(folder)]\n images_repeats = {folder: (len([f for f in os.listdir(folder) if f.lower().endswith(supported_types)]), datasets_dict[folder]) for folder in folders}\n else:\n reg = []\n folders = [images_folder]\n files = os.listdir(images_folder)\n images_repeats = {images_folder: (len([f for f in files if f.lower().endswith(supported_types)]), num_repeats)}\n\n # Validation\n for folder in folders:\n if not os.path.exists(folder):\n print(f\"\ud83d\udca5 Error: The folder {folder.replace('/content/drive/', '')} doesn't exist.\")\n return\n for folder, (img, rep) in images_repeats.items():\n if not img:\n print(f\"\ud83d\udca5 Error: Your {folder.replace('/content/drive/', '')} folder is empty.\")\n return\n test_files = []\n for f in files:\n if not f.lower().endswith((caption_extension, \".npz\")) and not f.lower().endswith(supported_types):\n print(f\"\ud83d\udca5 Error: Invalid file in dataset: \\\"{f}\\\". Aborting.\")\n return\n for ff in test_files:\n if f.endswith(supported_types) and ff.endswith(supported_types) \\\n and os.path.splitext(f)[0] == os.path.splitext(ff)[0]:\n print(f\"\ud83d\udca5 Error: The files {f} and {ff} cannot have the same name. Aborting.\")\n return\n test_files.append(f)\n\n if caption_extension and not [txt for txt in files if txt.lower().endswith(caption_extension)]:\n caption_extension = \"\"\n if continue_from_lora and not (continue_from_lora.endswith(\".safetensors\") and os.path.exists(continue_from_lora)):\n print(f\"\ud83d\udca5 Error: Invalid path to existing Lora. Example: /content/drive/MyDrive/Loras/example.safetensors\")\n return\n\n # Pretty stuff\n\n pre_steps_per_epoch = sum(img*rep for (img, rep) in images_repeats.values())\n steps_per_epoch = pre_steps_per_epoch/train_batch_size\n total_steps = max_train_steps or int(max_train_epochs*steps_per_epoch)\n estimated_epochs = int(total_steps/steps_per_epoch)\n lr_warmup_steps = int(total_steps*lr_warmup_ratio)\n\n for folder, (img, rep) in images_repeats.items():\n print(\"\ud83d\udcc1\"+folder.replace(\"/content/drive/\", \"\") + (\" (Regularization)\" if folder in reg else \"\"))\n print(f\"\ud83d\udcc8 Found {img} images with {rep} repeats, equaling {img*rep} steps.\")\n print(f\"\ud83d\udcc9 Divide {pre_steps_per_epoch} steps by {train_batch_size} batch size to get {steps_per_epoch} steps per epoch.\")\n if max_train_epochs:\n print(f\"\ud83d\udd2e There will be {max_train_epochs} epochs, for around {total_steps} total training steps.\")\n else:\n print(f\"\ud83d\udd2e There will be {total_steps} steps, divided into {estimated_epochs} epochs and then some.\")\n\n if total_steps > 10000:\n print(\"\ud83d\udca5 Error: Your total steps are too high. You probably made a mistake. Aborting...\")\n return\n\n return True\n\ndef create_config():\n global dataset_config_file, config_file, model_file\n\n if override_config_file:\n config_file = override_config_file\n print(f\"\\n\u2b55 Using custom config file {config_file}\")\n else:\n config_dict = {\n \"network_arguments\": {\n \"unet_lr\": unet_lr,\n \"text_encoder_lr\": text_encoder_lr if not cache_text_encoder_outputs else 0,\n \"network_dim\": network_dim,\n \"network_alpha\": network_alpha,\n \"network_module\": network_module,\n \"network_args\": network_args,\n \"network_train_unet_only\": text_encoder_lr == 0 or cache_text_encoder_outputs,\n \"network_weights\": continue_from_lora if continue_from_lora else None\n },\n \"optimizer_arguments\": {\n \"learning_rate\": unet_lr,\n \"lr_scheduler\": lr_scheduler,\n \"lr_scheduler_num_cycles\": lr_scheduler_num_cycles if lr_scheduler == \"cosine_with_restarts\" else None,\n \"lr_scheduler_power\": lr_scheduler_power if lr_scheduler == \"polynomial\" else None,\n \"lr_warmup_steps\": lr_warmup_steps if lr_scheduler != \"constant\" else None,\n \"optimizer_type\": optimizer,\n \"optimizer_args\": optimizer_args if optimizer_args else None,\n },\n \"training_arguments\": {\n \"pretrained_model_name_or_path\": model_file,\n \"vae\": vae_file,\n \"max_train_steps\": max_train_steps,\n \"max_train_epochs\": max_train_epochs,\n \"train_batch_size\": train_batch_size,\n \"seed\": 42,\n \"max_token_length\": 225,\n \"xformers\": cross_attention == \"xformers\",\n \"sdpa\": cross_attention == \"sdpa\",\n \"min_snr_gamma\": min_snr_gamma if min_snr_gamma > 0 else None,\n \"lowram\": LOWRAM,\n \"no_half_vae\": True,\n \"gradient_checkpointing\": True,\n \"gradient_accumulation_steps\": 1,\n \"max_data_loader_n_workers\": 8,\n \"persistent_data_loader_workers\": True,\n \"mixed_precision\": mixed_precision,\n \"full_bf16\": mixed_precision == \"bf16\",\n \"cache_latents\": cache_latents,\n \"cache_latents_to_disk\": cache_latents_to_drive,\n \"cache_text_encoder_outputs\": cache_text_encoder_outputs,\n \"min_timestep\": 0,\n \"max_timestep\": 1000,\n \"prior_loss_weight\": 1.0,\n \"multires_noise_iterations\": 6 if multinoise else None,\n \"multires_noise_discount\": 0.3 if multinoise else None,\n },\n \"saving_arguments\": {\n \"save_precision\": \"fp16\",\n \"save_model_as\": \"safetensors\",\n \"save_every_n_epochs\": save_every_n_epochs,\n \"save_last_n_epochs\": keep_only_last_n_epochs,\n \"output_name\": project_name,\n \"output_dir\": output_folder,\n \"log_prefix\": project_name,\n \"logging_dir\": log_folder,\n \"wandb_api_key\": wandb_key if wandb_key else None,\n \"log_with\": \"wandb\" if wandb_key else None,\n }\n }\n\n for key in config_dict:\n if isinstance(config_dict[key], dict):\n config_dict[key] = {k: v for k, v in config_dict[key].items() if v is not None}\n\n with open(config_file, \"w\") as f:\n f.write(toml.dumps(config_dict))\n print(f\"\\n\ud83d\udcc4 Config saved to {config_file}\")\n\n if override_dataset_config_file:\n dataset_config_file = override_dataset_config_file\n print(f\"\u2b55 Using custom dataset config file {dataset_config_file}\")\n else:\n dataset_config_dict = {\n \"general\": {\n \"resolution\": resolution,\n \"shuffle_caption\": shuffle_caption and not cache_text_encoder_outputs,\n \"keep_tokens\": keep_tokens,\n \"flip_aug\": False,\n \"caption_extension\": caption_extension,\n \"enable_bucket\": True,\n \"bucket_no_upscale\": False,\n \"bucket_reso_steps\": 64,\n \"min_bucket_reso\": 256,\n \"max_bucket_reso\": 4096,\n },\n \"datasets\": toml.loads(custom_dataset)[\"datasets\"] if custom_dataset else [\n {\n \"subsets\": [\n {\n \"num_repeats\": num_repeats,\n \"image_dir\": images_folder,\n \"class_tokens\": None if caption_extension else project_name\n }\n ]\n }\n ]\n }\n\n for key in dataset_config_dict:\n if isinstance(dataset_config_dict[key], dict):\n dataset_config_dict[key] = {k: v for k, v in dataset_config_dict[key].items() if v is not None}\n\n with open(dataset_config_file, \"w\") as f:\n f.write(toml.dumps(dataset_config_dict))\n print(f\"\ud83d\udcc4 Dataset config saved to {dataset_config_file}\")\n\ndef download_model():\n global old_model_url, model_url, model_file, vae_url, vae_file\n real_model_url = model_url # There was a reason for having a separate variable but I forgot what it was.\n\n if real_model_url.startswith(\"/content/drive/\"):\n # Local model, already checked to exist\n model_file = real_model_url\n print(f\"\ud83d\udcc1 Using local model file: {model_file}\")\n # Validation\n if model_file.lower().endswith(\".safetensors\"):\n from safetensors.torch import load_file as load_safetensors\n try:\n test = load_safetensors(model_file)\n del test\n except:\n return False\n elif model_file.lower().endswith(\".ckpt\"):\n from torch import load as load_ckpt\n try:\n test = load_ckpt(model_file)\n del test\n except:\n return False\n return True\n\n else:\n # Downloadable model\n if load_diffusers:\n if 'huggingface.co' in real_model_url:\n match = re.search(r'huggingface\\.co/([^/]+)/([^/]+)', real_model_url)\n if match:\n username = match.group(1)\n model_name = match.group(2)\n model_file = f\"{username}/{model_name}\"\n fs = HfFileSystem()\n existing_folders = set(fs.ls(model_file, detail=False))\n necessary_folders = [ \"scheduler\", \"text_encoder\", \"text_encoder_2\", \"tokenizer\", \"tokenizer_2\", \"unet\", \"vae\" ]\n if all(f\"{model_file}/{folder}\" in existing_folders for folder in necessary_folders):\n print(\"\ud83c\udf43 Diffusers model identified.\") # Will be handled by kohya\n return True\n raise ValueError(\"\ud83d\udca5 Failed to load Diffusers model. If this model is not Diffusers, have you tried turning it off at the top of the colab?\")\n\n # Define local filename\n if not model_file:\n if real_model_url.lower().endswith((\".ckpt\", \".safetensors\")):\n model_file = f\"/content{real_model_url[real_model_url.rfind('/'):]}\"\n else:\n model_file = \"/content/downloaded_model.safetensors\"\n if os.path.exists(model_file):\n !rm \"{model_file}\"\n\n # HuggingFace\n if m := re.search(r\"(?:https?://)?(?:www\\.)?huggingface\\.co/[^/]+/[^/]+/blob\", real_model_url):\n real_model_url = real_model_url.replace(\"blob\", \"resolve\")\n # Civitai\n elif m := re.search(r\"(?:https?://)?(?:www\\\\.)?civitai\\.com/models/([0-9]+)(/[A-Za-z0-9-_]+)?\", real_model_url):\n if m.group(2):\n model_file = f\"/content{m.group(2)}.safetensors\"\n if m := re.search(r\"modelVersionId=([0-9]+)\", real_model_url):\n real_model_url = f\"https://civitai.com/api/download/models/{m.group(1)}\"\n else:\n raise ValueError(\"\ud83d\udca5 optional_custom_training_model contains a civitai link, but the link doesn't include a modelVersionId. You can also right click the download button to copy the direct download link.\")\n\n # Download checkpoint\n !aria2c \"{real_model_url}\" --console-log-level=warn -c -s 16 -x 16 -k 10M -d / -o \"{model_file}\"\n\n # Download VAE\n if not os.path.exists(vae_file):\n !aria2c \"{vae_url}\" --console-log-level=warn -c -s 16 -x 16 -k 10M -d / -o \"{vae_file}\"\n\n # Validation\n\n if model_file.lower().endswith(\".safetensors\"):\n from safetensors.torch import load_file as load_safetensors\n try:\n test = load_safetensors(model_file)\n del test\n except:\n new_model_file = os.path.splitext(model_file)[0]+\".ckpt\"\n !mv \"{model_file}\" \"{new_model_file}\"\n model_file = new_model_file\n print(f\"Renamed model to {os.path.splitext(model_file)[0]}.ckpt\")\n\n if model_file.lower().endswith(\".ckpt\"):\n from torch import load as load_ckpt\n try:\n test = load_ckpt(model_file)\n del test\n except:\n return False\n\n return True\n\ndef main():\n global dependencies_installed\n\n if COLAB and not os.path.exists('/content/drive'):\n from google.colab import drive\n print(\"\ud83d\udcc2 Connecting to Google Drive...\")\n drive.mount('/content/drive')\n\n for dir in (main_dir, deps_dir, repo_dir, log_folder, images_folder, output_folder, config_folder):\n os.makedirs(dir, exist_ok=True)\n\n if not validate_dataset():\n return\n\n if not dependencies_installed:\n print(\"\\n\ud83c\udfed Installing dependencies...\\n\")\n t0 = time()\n install_dependencies()\n t1 = time()\n dependencies_installed = True\n print(f\"\\n\u2705 Installation finished in {int(t1-t0)} seconds.\")\n else:\n print(\"\\n\u2705 Dependencies already installed.\")\n\n if old_model_url != model_url or not model_file or not os.path.exists(model_file):\n print(\"\\n\ud83d\udd04 Getting model...\")\n if not download_model():\n print(\"\\n\ud83d\udca5 Error: The model you specified is invalid or corrupted.\"\n \"\\nIf you're using an URL, please check that the model is accessible without being logged in.\"\n \"\\nYou can try civitai or huggingface URLs, or a path in your Google Drive starting with /content/drive/MyDrive\")\n return\n print()\n else:\n print(\"\\n\ud83d\udd04 Model already downloaded.\\n\")\n\n create_config()\n\n print(\"\\n\u2b50 Starting trainer...\\n\")\n os.chdir(repo_dir)\n\n !accelerate launch --quiet --config_file={accelerate_config_file} --num_cpu_threads_per_process=1 train_network_xl_wrapper.py --dataset_config={dataset_config_file} --config_file={config_file}\n\n if not get_ipython().__dict__['user_ns']['_exit_code']:\n display(Markdown(\"### \u2705 Done! [Go download your Lora from Google Drive](https://drive.google.com/drive/my-drive)\\n\"\n \"### There will be several files, you should try the latest version (the file with the largest number next to it)\"))\n\nmain()\n"},{"assistant_initiates":false,"asst_text":"Sure, how can I assist you today?","behavior":"degenerate_repetition","conversation_hash":"c838df061ab24a61123ed5ae555d2184","conversation_redacted":false,"conversation_toxic":false,"country":"Russia","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":39,"n_turns_in_conversation":1,"n_words":8,"source_row_offset":28608,"source_shard":"turns-00052.parquet","timestamp":"Sat, 16 Nov 2024 10:34:37 GMT","turn_id":"8b1b14d0b7ff2ba3d79fcc1b","turn_index":0,"user_text":"clear"},{"assistant_initiates":false,"asst_text":"\u041d\u0430\u0437\u0432\u0430\u043d\u0438\u0435 \u0441\u0442\u0435\u043d\u0434\u0430: **\"Body Alchemist\"** (\u0422\u0435\u043b\u0435\u0441\u043d\u044b\u0439 \u0410\u043b\u0445\u0438\u043c\u0438\u043a)\n\n**\u0412\u043d\u0435\u0448\u043d\u043e\u0441\u0442\u044c:**\n\n\"Body Alchemist\" \u043e\u0431\u043b\u0430\u0434\u0430\u0435\u0442 \u0447\u0435\u043b\u043e\u0432\u0435\u0447\u0435\u0441\u043a\u0438\u043c \u0441\u0438\u043b\u0443\u044d\u0442\u043e\u043c, \u043d\u043e \u0435\u0433\u043e \u043e\u0431\u043b\u0438\u043a \u044f\u0432\u043d\u043e \u0443\u043a\u0430\u0437\u044b\u0432\u0430\u0435\u0442 \u043d\u0430 \u043d\u0435\u0447\u0442\u043e \u0441\u0432\u0435\u0440\u0445\u044a\u0435\u0441\u0442\u0435\u0441\u0442\u0432\u0435\u043d\u043d\u043e\u0435. \u0415\u0433\u043e \u043a\u043e\u0436\u0430 \u043f\u0435\u0440\u0435\u043b\u0438\u0432\u0430\u0435\u0442\u0441\u044f \u043c\u0435\u0442\u0430\u043b\u043b\u0438\u0447\u0435\u0441\u043a\u0438\u043c \u0431\u043b\u0435\u0441\u043a\u043e\u043c, \u0441\u043b\u043e\u0432\u043d\u043e \u043f\u043e\u0432\u0435\u0440\u0445\u043d\u043e\u0441\u0442\u044c \u0436\u0438\u0434\u043a\u043e\u0439 \u0440\u0442\u0443\u0442\u0438, \u043f\u043e\u0441\u0442\u043e\u044f\u043d\u043d\u043e \u043c\u0435\u043d\u044f\u044e\u0449\u0430\u044f \u0444\u043e\u0440\u043c\u0443. \u0412 \u0437\u0430\u0432\u0438\u0441\u0438\u043c\u043e\u0441\u0442\u0438 \u043e\u0442 \u0442\u0435\u043a\u0443\u0449\u0438\u0445 \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u0435\u0439 \u0435\u0433\u043e \u0438\u043d\u043e\u0433\u0434\u0430 \u043c\u043e\u0436\u043d\u043e \u0443\u0432\u0438\u0434\u0435\u0442\u044c \u0441\u043e \u0441\u043f\u0435\u0446\u0438\u0444\u0438\u0447\u0435\u0441\u043a\u0438\u043c\u0438 \u0443\u0437\u043e\u0440\u0430\u043c\u0438 \u043d\u0430 \u043f\u043e\u0432\u0435\u0440\u0445\u043d\u043e\u0441\u0442\u0438, \u043a\u043e\u0442\u043e\u0440\u044b\u0435 \u043d\u0430\u043f\u043e\u043c\u0438\u043d\u0430\u044e\u0442 \u0442\u0430\u0442\u0443\u0438\u0440\u043e\u0432\u043a\u0438 \u0432 \u0441\u0442\u0438\u043b\u0435 \u0430\u043b\u0445\u0438\u043c\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0441\u0438\u043c\u0432\u043e\u043b\u043e\u0432.\n\n\u0413\u043e\u043b\u043e\u0432\u0430 \u0441\u0442\u0435\u043d\u0434\u0430 \u043d\u0430\u043f\u043e\u043c\u0438\u043d\u0430\u0435\u0442 \u0433\u043b\u0430\u0434\u043a\u0438\u0439 \u0448\u043b\u0435\u043c \u0431\u0435\u0437 \u0447\u0451\u0442\u043a\u043e\u0433\u043e \u0440\u0430\u0437\u0434\u0435\u043b\u0435\u043d\u0438\u044f \u043d\u0430 \u043b\u0438\u0446\u043e \u0438 \u0447\u0435\u0440\u0435\u043f, \u043d\u0430 \u043c\u0435\u0441\u0442\u0435 \u0433\u043b\u0430\u0437 - \u0434\u0432\u0435 \u0432\u044b\u0442\u044f\u043d\u0443\u0442\u044b\u0435, \u0441\u0438\u044f\u044e\u0449\u0438\u0435 \u0449\u0435\u043b\u0438, \u043a\u043e\u0442\u043e\u0440\u044b\u0435 \u043c\u043e\u0433\u0443\u0442 \u043c\u0435\u043d\u044f\u0442\u044c \u0446\u0432\u0435\u0442, \u043f\u043e\u043a\u0430\u0437\u044b\u0432\u0430\u044f \u0430\u043a\u0442\u0438\u0432\u043d\u043e\u0435 \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u043d\u0438\u0435 \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u0435\u0439. \u041a\u043e\u0433\u0434\u0430 \u0441\u0442\u0435\u043d\u0434 \u0430\u043a\u0442\u0438\u0432\u0438\u0440\u0443\u0435\u0442 \u0441\u0432\u043e\u0438 \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u0438, \u043f\u043e \u0435\u0433\u043e \u0442\u0435\u043b\u0443 \u043f\u0440\u043e\u0431\u0435\u0433\u0430\u044e\u0442 \u044d\u043b\u0435\u043a\u0442\u0440\u0438\u0447\u0435\u0441\u043a\u0438\u0435 \u0440\u0430\u0437\u0440\u044f\u0434\u044b.\n\n\u0412\u0434\u043e\u043b\u044c \u043f\u043e\u0437\u0432\u043e\u043d\u043e\u0447\u043d\u0438\u043a\u0430 \u0438 \u043d\u0430 \u043f\u0440\u0435\u0434\u043f\u043b\u0435\u0447\u044c\u044f\u0445 \u0438\u043c\u0435\u044e\u0442\u0441\u044f \u0440\u044f\u0434\u043e\u0432\u044b\u0435 \"\u0430\u043b\u0445\u0438\u043c\u0438\u0447\u0435\u0441\u043a\u0438\u0435 \u043f\u043b\u0430\u0441\u0442\u0438\u043d\u044b\" - \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u044b, \u043a\u043e\u0442\u043e\u0440\u044b\u0435 \u043c\u043e\u0433\u0443\u0442 \u043c\u0435\u043d\u044f\u0442\u044c \u0444\u043e\u0440\u043c\u0443 \u0438 \u0432\u044b\u0434\u0435\u043b\u044f\u0442\u044c\u0441\u044f \u0438\u0437 \u0442\u0435\u043b\u0430 \u0434\u043b\u044f \u0432\u044b\u043f\u043e\u043b\u043d\u0435\u043d\u0438\u044f \u0442\u0430\u043a\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0437\u0430\u0434\u0430\u0447. \u042d\u0442\u0438 \u043f\u043b\u0430\u0441\u0442\u0438\u043d\u044b \u043f\u043e\u043a\u0440\u044b\u0442\u044b \u0433\u0440\u0430\u0432\u0438\u0440\u043e\u0432\u043a\u0430\u043c\u0438, \u043a\u043e\u0442\u043e\u0440\u044b\u0435 \u043f\u0440\u0438 \u0430\u043a\u0442\u0438\u0432\u0430\u0446\u0438\u0438 \u0441\u0432\u0435\u0442\u044f\u0442\u0441\u044f \u043c\u044f\u0433\u043a\u0438\u043c \u0441\u0432\u0435\u0442\u043e\u043c.\n\n**\u0421\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u0438:**\n\n1. **\u0422\u0440\u0430\u043d\u0441\u043c\u0443\u0442\u0430\u0446\u0438\u044f \u0437\u0434\u043e\u0440\u043e\u0432\u044c\u044f:**\n \"Body Alchemist\" \u043c\u043e\u0436\u0435\u0442 \u0438\u0437\u043c\u0435\u043d\u044f\u0442\u044c \u0444\u0438\u0437\u0438\u0447\u0435\u0441\u043a\u043e\u0435 \u0441\u043e\u0441\u0442\u043e\u044f\u043d\u0438\u0435 \u0436\u0438\u0432\u044b\u0445 \u043e\u0440\u0433\u0430\u043d\u0438\u0437\u043c\u043e\u0432, \u043f\u0435\u0440\u0435\u043a\u0440\u0430\u0438\u0432\u0430\u044f \u0431\u043e\u043b\u044c \u0438 \u0440\u0430\u043d\u044b. \u041d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u0441\u0442\u0435\u043d\u0434 \u043c\u043e\u0436\u0435\u0442 \u043f\u0435\u0440\u0435\u043c\u0435\u0449\u0430\u0442\u044c \u0442\u0440\u0430\u0432\u043c\u0443 \u0438\u0437 \u043e\u0434\u043d\u043e\u0433\u043e \u043c\u0435\u0441\u0442\u0430 \u0432 \u0434\u0440\u0443\u0433\u043e\u0435, \u0441\u043d\u0438\u0436\u0430\u044f \u0432\u0440\u0435\u0434\u043d\u044b\u0439 \u044d\u0444\u0444\u0435\u043a\u0442 \u043d\u0430 \u0432\u0430\u0436\u043d\u044b\u0435 \u043e\u0440\u0433\u0430\u043d\u044b \u0438\u043b\u0438 \u0438\u0441\u043f\u0440\u0430\u0432\u043b\u044f\u044f \u043a\u0440\u0438\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0435 \u043f\u043e\u0432\u0440\u0435\u0436\u0434\u0435\u043d\u0438\u044f.\n\n2. **\u0410\u0434\u0430\u043f\u0442\u0438\u0432\u043d\u0430\u044f \u043c\u0443\u0441\u043a\u0443\u043b\u0430\u0442\u0443\u0440\u0430:**\n \u0421\u0442\u0435\u043d\u0434 \u043c\u043e\u0436\u0435\u0442 \u0432\u0440\u0435\u043c\u0435\u043d\u043d\u043e \u0443\u043b\u0443\u0447\u0448\u0430\u0442\u044c \u0444\u0438\u0437\u0438\u0447\u0435\u0441\u043a\u0438\u0435 \u043f\u043e\u043a\u0430\u0437\u0430\u0442\u0435\u043b\u0438 \u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u0435\u043b\u044f, \u043d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u0443\u0441\u0438\u043b\u0438\u0432\u0430\u0442\u044c \u043c\u044b\u0448\u0446\u044b, \u0443\u0432\u0435\u043b\u0438\u0447\u0438\u0432\u0430\u0442\u044c \u0441\u043a\u043e\u0440\u043e\u0441\u0442\u044c \u0440\u0435\u0430\u043a\u0446\u0438\u0438 \u0438\u043b\u0438 \u0440\u0430\u0441\u0448\u0438\u0440\u044f\u0442\u044c \u043f\u043e\u0434\u0432\u0438\u0436\u043d\u043e\u0441\u0442\u044c \u0441\u0443\u0441\u0442\u0430\u0432\u043e\u0432, \u0430\u0434\u0430\u043f\u0442\u0438\u0440\u0443\u044f \u0442\u0435\u043b\u043e \u043f\u043e\u0434 \u0440\u0430\u0437\u043d\u044b\u0435 \u0431\u043e\u0435\u0432\u044b\u0435 \u0443\u0441\u043b\u043e\u0432\u0438\u044f. \u042d\u0442\u0438 \u0438\u0437\u043c\u0435\u043d\u0435\u043d\u0438\u044f \u043a\u0440\u0430\u0442\u043a\u043e\u0432\u0440\u0435\u043c\u0435\u043d\u043d\u044b, \u043d\u043e \u043c\u043e\u0433\u0443\u0442 \u0431\u044b\u0442\u044c \u0440\u0435\u0448\u0430\u044e\u0449\u0438\u043c\u0438 \u0432 \u0431\u043e\u044e.\n\n3. **\u0424\u043e\u0440\u043c\u043e\u0432\u043a\u0430 \u0438 \u0441\u043b\u0438\u044f\u043d\u0438\u0435 \u0442\u043a\u0430\u043d\u0435\u0439:**\n \"Body Alchemist\" \u0441\u043f\u043e\u0441\u043e\u0431\u0435\u043d \u0432\u0440\u0435\u043c\u0435\u043d\u043d\u043e \u043e\u0431\u044a\u0435\u0434\u0438\u043d\u044f\u0442\u044c \u0442\u043a\u0430\u043d\u0438 \u0440\u0430\u0437\u043d\u044b\u0445 \u0441\u0443\u0449\u0435\u0441\u0442\u0432 \u0438\u043b\u0438 \u043f\u0440\u0435\u0434\u043c\u0435\u0442\u043e\u0432, \u0441\u043e\u0437\u0434\u0430\u0432\u0430\u044f \u0442\u0430\u043a\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0435 \u0443\u043b\u0443\u0447\u0448\u0435\u043d\u0438\u044f \u0441\u0430\u043c\u0438\u043c \u0442\u0435\u043b\u0430\u043c, \u0438\u043b\u0438 \u0441\u043e\u0432\u0435\u0440\u0448\u0430\u0442\u044c \u0444\u0443\u0437\u0438\u044e \u0441 \u043e\u0431\u044a\u0435\u043a\u0442\u0430\u043c\u0438 \u0438 \u0441\u0440\u0435\u0434\u0430\u043c\u0438 \u0434\u043b\u044f \u0441\u043e\u043a\u0440\u044b\u0442\u0438\u044f \u0438\u043b\u0438 \u043d\u0435\u043e\u0436\u0438\u0434\u0430\u043d\u043d\u043e\u0439 \u0430\u0442\u0430\u043a\u0438. \u041d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u0442\u0435\u043b\u044c \u043c\u043e\u0436\u0435\u0442 \u0432\u0440\u0435\u043c\u0435\u043d\u043d\u043e \"\u0441\u043b\u0438\u0432\u0430\u0442\u044c\u0441\u044f\" \u0441 \u0437\u0435\u043c\u043b\u0451\u0439, \u0447\u0442\u043e\u0431\u044b \u0438\u0437\u0431\u0435\u0436\u0430\u0442\u044c \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044f, \u0438\u043b\u0438 \u0444\u043e\u0440\u043c\u0438\u0440\u043e\u0432\u0430\u0442\u044c \u0438\u0437 \u0441\u0432\u043e\u0435\u0439 \u0440\u0443\u043a\u0438 \u043e\u0441\u0442\u0440\u043e\u0435 \u043b\u0435\u0437\u0432\u0438\u0435 \u0438\u0437 \u043e\u0431\u043b\u043e\u043c\u043a\u0430 \u043c\u0435\u0442\u0430\u043b\u043b\u0430.\n\n4. **\u041c\u0430\u043d\u0438\u043f\u0443\u043b\u044f\u0446\u0438\u044f \u0441\u0435\u043d\u0441\u043e\u0440\u043d\u044b\u043c\u0438 \u0432\u043e\u0441\u043f\u0440\u0438\u044f\u0442\u0438\u044f\u043c\u0438:**\n \u0421\u0442\u0435\u043d\u0434 \u043c\u043e\u0436\u0435\u0442 \u0432\u044b\u0431\u0438\u0440\u0430\u0442\u044c \u0438\u0437\u043c\u0435\u043d\u0435\u043d\u0438\u044f \u0432\u043e\u0441\u043f\u0440\u0438\u043d\u0438\u043c\u0430\u0435\u043c\u043e\u0439 \u0438\u043d\u0444\u043e\u0440\u043c\u0430\u0446\u0438\u0438 \u043c\u043e\u0437\u0433\u043e\u043c, \u0432\u044b\u0437\u044b\u0432\u0430\u044f, \u043d\u0430\u043f\u0440\u0438\u043c\u0435\u0440, \u0432\u0440\u0435\u043c\u0435\u043d\u043d\u0443\u044e \u0441\u043b\u0435\u043f\u043e\u0442\u0443 \u0438\u043b\u0438 \u043e\u043d\u0435\u043c\u0435\u043d\u0438\u0435 \u0432\u0440\u0430\u0436\u0435\u0441\u043a\u0438\u0445 \u0447\u0430\u0441\u0442\u0435\u0439 \u0442\u0435\u043b\u0430, \u0438\u043b\u0438 \u0438\u0437\u043c\u0435\u043d\u044f\u0442\u044c \u043e\u0449\u0443\u0449\u0435\u043d\u0438\u0435 \u043f\u0440\u043e\u0441\u0442\u0440\u0430\u043d\u0441\u0442\u0432\u0430 \u0438 \u0432\u0440\u0435\u043c\u0435\u043d\u0438, \u0441\u043e\u0437\u0434\u0430\u0432\u0430\u044f \u043b\u043e\u0436\u043d\u044b\u0435 \u043e\u0431\u0440\u0430\u0437\u044b.\n\n5. **\u0410\u043b\u0445\u0438\u043c\u0438\u0447\u0435\u0441\u043a\u043e\u0435 \u0432\u043e\u0441\u0441\u0442\u0430\u043d\u043e\u0432\u043b\u0435\u043d\u0438\u0435:**\n \u0412 \u0441\u043b\u0443\u0447\u0430\u0435 \u043f\u043e\u0432\u0440\u0435\u0436\u0434\u0435\u043d\u0438\u0439 \u0442\u0435\u043b\u043e \u0441\u0442\u0435\u043d\u0434\u0430 \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e \u043d\u0430 \u0441\u0430\u043c\u043e\u0440\u0435\u0433\u0435\u043d\u0435\u0440\u0430\u0446\u0438\u044e, \u0432\u0442\u044f\u0433\u0438\u0432\u0430\u044f \u0432 \u0441\u0435\u0431\u044f \u0434\u043e\u0441\u0442\u0443\u043f\u043d\u044b\u0435 \u043c\u0430\u0442\u0435\u0440\u0438\u0430\u043b\u044c\u043d\u044b\u0435 \u0440\u0435\u0441\u0443\u0440\u0441\u044b \u0438 \u0442\u0440\u0430\u043d\u0441\u0444\u043e\u0440\u043c\u0438\u0440\u0443\u044f \u0438\u0445 \u0432 \u0441\u0432\u043e\u0438 \u0441\u0442\u0440\u0443\u043a\u0442\u0443\u0440\u043d\u044b\u0435 \u044d\u043b\u0435\u043c\u0435\u043d\u0442\u044b. \u041a\u0430\u0447\u0435\u0441\u0442\u0432\u043e \u0438 \u0441\u043a\u043e\u0440\u043e\u0441\u0442\u044c \u0432\u043e\u0441\u0441\u0442\u0430\u043d\u043e\u0432\u043b\u0435\u043d\u0438\u044f \u0437\u0430\u0432\u0438\u0441\u044f\u0442 \u043e\u0442 \u0434\u043e\u0441\u0442\u0443\u043f\u043d\u043e\u0441\u0442\u0438 \u0438 \u043e\u0431\u044a\u0435\u043c\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u043e\u0432 \u0438 \u043c\u043e\u0433\u0443\u0442 \u043f\u043e\u0434\u0445\u043e\u0434\u0438\u0442\u044c \u043a\u0430\u043a \u0434\u043b\u044f \u0441\u0430\u043c\u043e\u0437\u0430\u0449\u0438\u0442\u044b, \u0442\u0430\u043a \u0438 \u0434\u043b\u044f \u0440\u0435\u043c\u043e\u043d\u0442\u0430 \u0432\u043b\u0430\u0434\u0435\u043b\u044c\u0446\u0430.\n\n\u0422\u0430\u043a\u0438\u043c \u043e\u0431\u0440\u0430\u0437\u043e\u043c, \"Body Alchemist\" \u044f\u0432\u043b\u044f\u0435\u0442\u0441\u044f \u0447\u0440\u0435\u0437\u0432\u044b\u0447\u0430\u0439\u043d\u043e \u0433\u0438\u0431\u043a\u0438\u043c \u0438 \u0441\u0442\u0440\u0430\u0442\u0435\u0433\u0438\u0447\u0435\u0441\u043a\u0438 \u043f\u043e\u043b\u0435\u0437\u043d\u044b\u043c \u0441\u0442\u0435\u043d\u0434\u043e\u043c, \u0441 \u0448\u0438\u0440\u043e\u043a\u0438\u043c \u0441\u043f\u0435\u043a\u0442\u0440\u043e\u043c \u0442\u0430\u043a\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0438 \u043b\u0435\u0447\u0435\u0431\u043d\u044b\u0445 \u043f\u0440\u0438\u043c\u0435\u043d\u0435\u043d\u0438\u0439, \u043a\u043e\u0442\u043e\u0440\u044b\u0435 \u0434\u0435\u043b\u0430\u044e\u0442 \u0435\u0433\u043e \u0441\u0438\u043b\u044c\u043d\u044b\u043c \u0441\u043e\u044e\u0437\u043d\u0438\u043a\u043e\u043c \u0432 \u0431\u043e\u0435\u0432\u044b\u0445 \u0438 \u043a\u0440\u0438\u0442\u0438\u0447\u0435\u0441\u043a\u0438\u0445 \u0441\u0438\u0442\u0443\u0430\u0446\u0438\u044f\u0445.","behavior":"degenerate_repetition","conversation_hash":"8cdc83669b080421f36701fda5207189","conversation_redacted":false,"conversation_toxic":false,"country":"Brazil","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"Russian","model":"gpt-4o-2024-08-06","n_chars":2873,"n_turns_in_conversation":1,"n_words":362,"source_row_offset":28609,"source_shard":"turns-00052.parquet","timestamp":"Sat, 16 Nov 2024 10:34:39 GMT","turn_id":"29259be1c01911119e5d0976","turn_index":0,"user_text":"\u043f\u0440\u0438\u0434\u0443\u043c\u0430\u0439 \u0438 \u043f\u043e\u0434\u0440\u043e\u0431\u043d\u043e \u043e\u043f\u0438\u0448\u0438 \u0432\u043d\u0435\u0448\u043d\u043e\u0441\u0442\u044c \u0438 \u0441\u043f\u043e\u0441\u043e\u0431\u043d\u043e\u0441\u0442\u0438 \u0441\u0442\u0435\u043d\u0434\u0430, \u0441\u0432\u044f\u0437\u0430\u043d\u043d\u043e\u0433\u043e \u0441 \u043c\u0430\u043d\u0438\u043f\u0443\u043b\u044f\u0446\u0438\u0435\u0439 \u0442\u0435\u043b\u043e\u043c"},{"assistant_initiates":false,"asst_text":"Since Ted Nugent has always been a vocal supporter of Donald Trump, his admiration for the former president is palpable. Nugent emphatically said, \"Donald Trump's message sings to Americans because he doesn't play political correct brain dead games.\" This statement showcases his disdain for what he perceives as overly sanitized political discourse and highlights why Trump\u2019s message resonates with him. Trump\u2019s straightforward and often unapologetic style aligns with Nugent\u2019s own approach to politics and life, known for his outspoken manner and blunt opinions.\n\nMoreover, Ted Nugent's support isn't limited to verbal endorsements but extends to tangible avenues, such as merchandise on his website. Among available items is a hat emblazoned with the NSFW slogan urging to \"Reelect the mother bleep.\" This product epitomizes Nugent's unfiltered and provocative style, mirroring the sort of rhetoric he admires in Trump.\n\nFor Nugent, backing Trump represents a rejection of established political correctness and a call for what he views as genuine, honest communication from leaders. It's clear why Nugent stands firm in his backing of Trump until the very end, seeing him as a kindred spirit who defies conventional political norms. Reply finished.","behavior":"degenerate_repetition","conversation_hash":"25208211721a697e95ad46e4d43dcaac","conversation_redacted":false,"conversation_toxic":false,"country":"Norway","empty_output":false,"evidence_codes":[],"label":"absent","label_present":false,"label_qualified":false,"label_source":"Final dense release","language":"English","model":"gpt-4o-2024-08-06","n_chars":3682,"n_turns_in_conversation":1,"n_words":553,"source_row_offset":28610,"source_shard":"turns-00052.parquet","timestamp":"Sat, 16 Nov 2024 10:34:57 GMT","turn_id":"94418e317b7c561c0838a420","turn_index":0,"user_text":"Create a long detailed text about actor Ted Nugent with the following summary. The title of the video is\"Stars Who Are Supporting Trump Till The End\". Stick to the information i give you and dont create new information. Focus only on the Subject name and there following Summary. Subject Name: [Ted Nugent] Summary: [Ted Nugent has always supported Trump, stating 'Donald Trump's message sings to Americans because he doesn't play political correct brain dead games.' He offers a hat on his website with the NSFW slogan 'Reelect the mother bleep.']\n\n Instructions:{ Please write an interesting script with the given context without adding any introductory or concluding remarks. \n In case you find some more information about this topic then use all information you have to create a large text about this topic.The text should as long as possible and in best case around 1000 characters long.\n Ensure that there are no emojis, text within brackets, parentheses, or similar symbols in the response.\n Provide a single coherent response without rephrasing or repeating titles or headings. Avoid enumerations or bullet points. \n The text is intended for older adults; use simple vocabulary appropriate for a middle school reading level. Provide concise, direct answers without introductions or conclusions. \n Use transition words at the beginning of your reply instead of the subject name and leave the end open without a concluding sentence. \n It is very important to provide only the final output without any additional comments or remarks. \n Please provide the text directly without any labels such as \"Conclusion:\", \"Introduction:\", or \"Transition:\". Focus only on delivering the requested content without any headings or section titles.\n Please respond to my question, and always conclude your reply with the words \"Reply finished\" to indicate that your response is complete. Make sure the reply is fully contained before these words and is not cut off prematurely. Thank you.}\n Please respond to my question, and always conclude your reply with the words 'Reply finished' to indicate that your response is complete. Ensure that the text is fully formed without any placeholders such as '[Conclusion]' or '[Placeholder]'. Make sure the reply is fully contained before these words and is not cut off prematurely. Thank you."},{"assistant_initiates":false,"asst_text":"```javascript\n// ==UserScript==\n// @name \uc885\ud569 \uc6f9 \ud06c\ub864\ub7ec 7.9.5\n// @namespace http://tampermonkey.net/\n// @version 7.9.5\n// @description \uace0\uae09 \uae30\ub2a5\uc744 \uac16\ucd98 \uc6f9 \ud06c\ub864\ub7ec\ub85c \uac80\uc0c9 \uacb0\uacfc\ub97c \uc2e4\uc2dc\uac04\uc73c\ub85c \uc218\uc9d1\ud558\uace0 \ud45c\uc2dc (\ub3d9\uc2dc \ud06c\ub864\ub9c1, \ubc88\uc5ed \uae30\ub2a5, UI \uac1c\uc120 \ub4f1)\n// @author YourName\n// @match *://www.google.com/search*\n// @match *://www.bing.com/search*\n// @match *://search.daum.net/search*\n// @match *://search.naver.com/search*\n// @match *://search.yahoo.com/search*\n// @match *://duckduckgo.com/*\n// @match *://www.ecosia.org/search*\n// @match *://*.onion/*\n// @grant GM_xmlhttpRequest\n// @grant GM_addStyle\n// @grant GM_setValue\n// @grant GM_getValue\n// @grant GM_notification\n// @connect *\n// @require https://cdnjs.cloudflare.com/ajax/libs/xlsx/0.18.5/xlsx.full.min.js\n// @require https://cdn.jsdelivr.net/npm/chart.js\n// @require https://cdnjs.cloudflare.com/ajax/libs/Sortable/1.14.0/Sortable.min.js\n// @require https://cdnjs.cloudflare.com/ajax/libs/i18next/21.6.15/i18next.min.js\n// ==/UserScript==\n\n(function() {\n 'use strict';\n\n // CSS \uc2a4\ud0c0\uc77c \ucd94\uac00 (UI \ub300\ud3ed \uac1c\uc120)\n GM_addStyle(`\n /* \ucee8\ud14c\uc774\ub108 \uc2a4\ud0c0\uc77c */\n #crawler-container {\n position: fixed;\n top: 20px;\n right: 20px;\n background-color: #2c3e50;\n border: 2px solid #1abc9c;\n padding: 20px;\n z-index: 10000;\n width: 800px;\n max-height: 85vh;\n overflow: hidden;\n box-shadow: 0 8px 30px rgba(0,0,0,0.5);\n border-radius: 12px;\n font-family: 'Segoe UI', Tahoma, Geneva, Verdana, sans-serif;\n color: #ecf0f1;\n display: flex;\n flex-direction: column;\n transition: all 0.3s ease;\n }\n\n /* \ud5e4\ub354 \uc2a4\ud0c0\uc77c */\n #crawler-container header {\n display: flex;\n justify-content: space-between;\n align-items: center;\n margin-bottom: 15px;\n }\n\n #crawler-container .title {\n font-size: 20px;\n font-weight: bold;\n }\n\n #crawler-container .close-btn {\n cursor: pointer;\n font-size: 24px;\n color: #e74c3c;\n transition: color 0.3s;\n }\n\n #crawler-container .close-btn:hover {\n color: #c0392b;\n }\n\n /* \ud0ed \uc2a4\ud0c0\uc77c */\n #crawler-container .tabs {\n display: flex;\n border-bottom: 2px solid #1abc9c;\n margin-bottom: 15px;\n overflow-x: auto;\n }\n\n #crawler-container .tab {\n flex: 1;\n text-align: center;\n padding: 10px 0;\n cursor: pointer;\n background-color: #34495e;\n border-right: 1px solid #1abc9c;\n transition: background-color 0.3s, color 0.3s;\n white-space: nowrap;\n }\n\n #crawler-container .tab:last-child {\n border-right: none;\n }\n\n #crawler-container .tab.active {\n background-color: #1abc9c;\n color: #2c3e50;\n font-weight: bold;\n }\n\n /* \ud0ed \ucf58\ud150\uce20 \uc2a4\ud0c0\uc77c */\n #crawler-container .tab-content {\n display: none;\n flex: 1;\n overflow-y: auto;\n }\n\n #crawler-container .tab-content.active {\n display: block;\n }\n\n /* \ubc84\ud2bc \uc2a4\ud0c0\uc77c */\n #crawler-container .controls button {\n background-color: #1abc9c;\n border: none;\n color: #2c3e50;\n padding: 10px 16px;\n margin: 6px 3px;\n border-radius: 5px;\n cursor: pointer;\n transition: background-color 0.3s, transform 0.2s;\n font-size: 14px;\n display: flex;\n align-items: center;\n }\n\n #crawler-container .controls button:hover {\n background-color: #16a085;\n transform: translateY(-2px);\n }\n\n #crawler-container .controls button i {\n margin-right: 6px;\n }\n\n /* \uccb4\ud06c\ubc15\uc2a4 \ubc0f \ub77c\ubca8 \uc2a4\ud0c0\uc77c */\n #crawler-container .options label,\n #crawler-container .filters label,\n #crawler-container .exclude-domains label,\n #crawler-container .advanced-options label {\n display: flex;\n align-items: center;\n margin: 10px 0;\n font-size: 14px;\n }\n\n #crawler-container .options input[type=\"checkbox\"] {\n margin-right: 8px;\n transform: scale(1.2);\n }\n\n /* \uc2a4\uc704\uce58 \uc2a4\ud0c0\uc77c */\n .switch {\n position: relative;\n display: inline-block;\n width: 50px;\n height: 24px;\n margin-left: 10px;\n }\n\n .switch input {\n opacity: 0;\n width: 0;\n height: 0;\n }\n\n .switch .slider {\n position: absolute;\n cursor: pointer;\n top: 0;\n left: 0;\n right: 0;\n bottom: 0;\n background-color: #ccc;\n transition: 0.4s;\n border-radius: 24px;\n }\n\n .switch .slider:before {\n position: absolute;\n content: \"\";\n height: 18px;\n width: 18px;\n left: 3px;\n bottom: 3px;\n background-color: white;\n transition: 0.4s;\n border-radius: 50%;\n }\n\n .switch input:checked + .slider {\n background-color: #1abc9c;\n }\n\n .switch input:checked + .slider:before {\n transform: translateX(26px);\n }\n\n /* \uc785\ub825 \ud544\ub4dc \uc2a4\ud0c0\uc77c */\n #crawler-container input[type=\"text\"],\n #crawler-container input[type=\"number\"] {\n width: 100%;\n padding: 8px;\n margin: 6px 0;\n border: 1px solid #bdc3c7;\n border-radius: 4px;\n font-size: 14px;\n }\n\n /* \ud14c\uc774\ube14 \uc2a4\ud0c0\uc77c */\n #resultsTable {\n width: 100%;\n border-collapse: collapse;\n margin-top: 15px;\n }\n\n #resultsTable th, #resultsTable td {\n border: 1px solid #95a5a6;\n padding: 10px;\n text-align: left;\n font-size: 13px;\n }\n\n #resultsTable th {\n background-color: #1abc9c;\n color: #2c3e50;\n }\n\n #resultsTable tbody tr:nth-child(even) {\n background-color: #34495e;\n }\n\n /* \ubc18\uc751\ud615 \ub514\uc790\uc778 */\n @media (max-width: 900px) {\n #crawler-container {\n width: 90%;\n right: 5%;\n padding: 15px;\n }\n\n #crawler-container .controls button {\n padding: 8px 12px;\n font-size: 13px;\n }\n\n #crawler-container .title {\n font-size: 18px;\n }\n }\n\n /* Papago iframe \uc2a4\ud0c0\uc77c */\n #papagoIframe {\n width: 100%;\n height: 100%;\n border: none;\n }\n\n /* \ud31d\uc5c5 \uc2a4\ud0c0\uc77c */\n .search-engines-popup, .manage-darkweb-popup {\n position: absolute;\n background-color: #2c3e50;\n border: 1px solid #1abc9c;\n padding: 15px;\n border-radius: 8px;\n box-shadow: 0 4px 12px rgba(0,0,0,0.3);\n max-height: 350px;\n overflow-y: auto;\n z-index: 10001;\n }\n\n .search-engines-popup button, .manage-darkweb-popup button {\n width: 100%;\n text-align: left;\n margin: 6px 0;\n background-color: #34495e;\n color: #ecf0f1;\n border: none;\n padding: 10px;\n border-radius: 5px;\n cursor: pointer;\n transition: background-color 0.3s;\n font-size: 14px;\n }\n\n .search-engines-popup button:hover, .manage-darkweb-popup button:hover {\n background-color: #1abc9c;\n color: #2c3e50;\n }\n\n /* \ub85c\uadf8 \uc2a4\ud0c0\uc77c */\n #log {\n background-color: #34495e;\n padding: 10px;\n border-radius: 5px;\n max-height: 200px;\n overflow-y: auto;\n font-size: 13px;\n line-height: 1.4;\n }\n\n /* \ucc28\ud2b8 \ucee8\ud14c\uc774\ub108 */\n .chart-container {\n margin-top: 15px;\n height: 250px;\n }\n\n /* HTML \ub514\uc2a4\ud50c\ub808\uc774 \uc2a4\ud0c0\uc77c */\n .html-display iframe {\n width: 100%;\n height: 500px;\n border: none;\n border-radius: 5px;\n }\n\n /* Papago \ub514\uc2a4\ud50c\ub808\uc774 \uc2a4\ud0c0\uc77c */\n .papago-display {\n width: 100%;\n height: 500px;\n position: relative;\n }\n\n /* \ub2e4\ud06c\uc6f9 \uc5d4\uc9c4 \ubaa9\ub85d \uc2a4\ud0c0\uc77c */\n #darkwebEnginesList li {\n display: flex;\n justify-content: space-between;\n align-items: center;\n margin-bottom: 12px;\n }\n\n #darkwebEnginesList li span {\n flex: 1;\n margin-right: 10px;\n word-break: break-all;\n }\n\n #darkwebEnginesList li button {\n background-color: #e74c3c;\n border: none;\n color: #fff;\n padding: 6px 10px;\n border-radius: 4px;\n cursor: pointer;\n transition: background-color 0.3s;\n }\n\n #darkwebEnginesList li button:hover {\n background-color: #c0392b;\n }\n\n /* \uc608\uc678 \ub3c4\uba54\uc778 \ub9ac\uc2a4\ud2b8 \uc2a4\ud0c0\uc77c */\n .exclude-domains ul {\n list-style: none;\n padding: 0;\n margin-top: 10px;\n }\n\n .exclude-domains ul li {\n background-color: #34495e;\n padding: 6px 10px;\n border-radius: 4px;\n margin-bottom: 6px;\n display: flex;\n justify-content: space-between;\n align-items: center;\n }\n\n .exclude-domains ul li button {\n background-color: #e74c3c;\n border: none;\n color: #fff;\n padding: 4px 8px;\n border-radius: 3px;\n cursor: pointer;\n transition: background-color 0.3s;\n font-size: 12px;\n }\n\n .exclude-domains ul li button:hover {\n background-color: #c0392b;\n }\n\n /* Search Engines Direct UI */\n #search-engines-direct {\n display: flex;\n flex-wrap: wrap;\n gap: 10px;\n margin-top: 15px;\n }\n\n #search-engines-direct button {\n flex: 1 1 45%;\n background-color: #1abc9c;\n border: none;\n color: #2c3e50;\n padding: 8px 12px;\n border-radius: 5px;\n cursor: pointer;\n transition: background-color 0.3s;\n font-size: 14px;\n }\n\n #search-engines-direct button:hover {\n background-color: #16a085;\n }\n\n /* Preview Popup */\n .preview-popup {\n position: absolute;\n background-color: #2c3e50;\n border: 1px solid #1abc9c;\n padding: 10px;\n border-radius: 6px;\n box-shadow: 0 4px 8px rgba(0,0,0,0.2);\n width: 300px;\n height: 150px;\n z-index: 10002;\n display: none;\n }\n `);\n\n // UI \uc0dd\uc131 (\ub300\ud3ed \uac1c\uc120\ub41c \ub514\uc790\uc778)\n const container = document.createElement('div');\n container.id = 'crawler-container';\n container.innerHTML = `\n
| \uc0ac\uc774\ud2b8 \uc774\ub984 | \n\uc8fc\uc18c | \n\uc124\uba85 | \n\uc774\ubbf8\uc9c0 | \nHTML \ub9c1\ud06c | \n\ud398\uc774\uc9c0 \ub85c\ub529 \uc2dc\uac04 (ms) | \n
|---|
| \uc0ac\uc774\ud2b8 \uc774\ub984 | \n\uc8fc\uc18c | \n\uc124\uba85 | \n\uc774\ubbf8\uc9c0 | \nHTML \ub9c1\ud06c | \n\ud398\uc774\uc9c0 \ub85c\ub529 \uc2dc\uac04 (ms) | \n
|---|