Good call! Pretty sure it's the same, but I've uploaded the new one just in case!

I've uploaded each ControlNet model as Civitai "versions" - along the top, under the title. Each ControlNet model is 700mb (I've cut them down in size, the originals are over a gig). The one you're downloading is the Tile model, that's correct. You'll also have to download Depth, Canny, etc.

@theally did cutting down remove anything quality related?

@StereoNostalgic218 No, the preprocessor results are indistinguishable from the results of the full models.

Absolutely! I'm doing my best to monitor the Huggingface/Git and keep them up to date over here!

Works just fine for me on ControlNet version 1.1.173. Reference uses no model, so it's extremely doubtful that any of my models (which are just pruned versions of the original models) would have any effect on the preprocessor!

@theally you're right it must have been an add on because it works now... but what about the i2p2? there is no "prompt" box :DD

""This checkpoint includes a config file, download and place it along side the checkpoint.""

Over on the right, under "Files" there's the model, and the .yaml

@theally Правильно ли я понимаю, что в итоге в папке должно быть два файла - yaml и safetensors?

So if the YAML says openpose.YAML but the model says Pretrained_OpenPosev10.safetensors. I'd have to change the YAML to pretrained_OpenPosev10.YAML?

@skogsra correct, just name the yaml the same as the model.

Why aren't they named correctly to begin with??

@yzy78550529 Civitai renames (or did, until recently) all uploaded models by appending the version to the model name, but not the config files, leaving them out of sync.

here ` \sd-webui-aki-v4.1\extensions\sd-webui-controlnet-main\models`

@ForestChen Thank you very much!

the Yaml files are extremely simple and short config text files. You can very literally open them up and see with one glance that they have no malicious code

@hgfhfhfhfh YAML files are not tensorflow files. They are plaintext human-readable config files.

lmao

i laugh till bleed.

If I recall correctly, HED became Softedge, I THINK.

Yes, thanks bro, I found it.

On the first load, the models need to be setup and you get the following warnings:

WARNING:fvcore.common.config:Loading config ...\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\oneformer\configs/ade20k/oneformer_swin_large_IN21k_384_bs16_160k.yaml with yaml.unsafe_load. Your machine may be at risk if the file contains malicious content.

WARNING:fvcore.common.config:Loading config ...\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\oneformer\configs/ade20k\Base-ADE20K-UnifiedSegmentation.yaml with yaml.unsafe_load. Your machine may be at risk if the file contains malicious content.

I get the same error. The exact issue is not that there is malicious content, but that it could load malicious content if there is any in the .yaml files "Base-ADE20K-UnifiedSegmentation.yaml" and "oneformer_swin_large_IN21k_384_bs16_160k.yaml". This is because the creator of the package is using "load_yaml_with_base" with the argument "allow_unsafe" set to True in lines 37/46 of "config(dot)py" file and "yaml.unsafe_load" in lines 228/295 of "lazy(dot)py"

The files in question are located in:

.../stable-diffusion-webui\extensions\sd-webui\controlnet\annotator\oneformer\detectron2\config\

I assume this is because the extension/model is still in development and it is difficult to ensure the .yaml is loaded safely... not really sure, though.

Looking at the two files:

"Base-ADE20K-UnifiedSegmentation.yaml"

MODEL: BACKBONE: FREEZE_AT: 0 NAME: "build_resnet_backbone" WEIGHTS: "detectron2://ImageNetPretrained/torchvision/R-50.pkl" PIXEL_MEAN: [123.675, 116.280, 103.530] PIXEL_STD: [58.395, 57.120, 57.375] RESNETS: DEPTH: 50 STEM_TYPE: "basic" # not used STEM_OUT_CHANNELS: 64 STRIDE_IN_1X1: False OUT_FEATURES: ["res2", "res3", "res4", "res5"] # NORM: "SyncBN" RES5_MULTI_GRID: [1, 1, 1] # not used DATASETS: TRAIN: ("ade20k_panoptic_train",) TEST_PANOPTIC: ("ade20k_panoptic_val",) TEST_INSTANCE: ("ade20k_instance_val",) TEST_SEMANTIC: ("ade20k_sem_seg_val",) SOLVER: IMS_PER_BATCH: 16 BASE_LR: 0.0001 MAX_ITER: 160000 WARMUP_FACTOR: 1.0 WARMUP_ITERS: 0 WEIGHT_DECAY: 0.05 OPTIMIZER: "ADAMW" LR_SCHEDULER_NAME: "WarmupPolyLR" BACKBONE_MULTIPLIER: 0.1 CLIP_GRADIENTS: ENABLED: True CLIP_TYPE: "full_model" CLIP_VALUE: 0.01 NORM_TYPE: 2.0 AMP: ENABLED: True INPUT: MIN_SIZE_TRAIN: !!python/object/apply:eval ["[int(x * 0.1 * 512) for x in range(5, 21)]"] MIN_SIZE_TRAIN_SAMPLING: "choice" MIN_SIZE_TEST: 512 MAX_SIZE_TRAIN: 2048 MAX_SIZE_TEST: 2048 CROP: ENABLED: True TYPE: "absolute" SIZE: (512, 512) SINGLE_CATEGORY_MAX_AREA: 1.0 COLOR_AUG_SSD: True SIZE_DIVISIBILITY: 512 # used in dataset mapper FORMAT: "RGB" DATASET_MAPPER_NAME: "oneformer_unified" MAX_SEQ_LEN: 77 TASK_SEQ_LEN: 77 TASK_PROB: SEMANTIC: 0.33 INSTANCE: 0.66 TEST: EVAL_PERIOD: 5000 AUG: ENABLED: False MIN_SIZES: [256, 384, 512, 640, 768, 896] MAX_SIZE: 3584 FLIP: True DATALOADER: FILTER_EMPTY_ANNOTATIONS: True NUM_WORKERS: 4 VERSION: 2

"oneformer_swin_large_IN21k_384_bs16_160k.yaml"

_BASE_: oneformer_R50_bs16_160k.yaml MODEL: BACKBONE: NAME: "D2SwinTransformer" SWIN: EMBED_DIM: 192 DEPTHS: [2, 2, 18, 2] NUM_HEADS: [6, 12, 24, 48] WINDOW_SIZE: 12 APE: False DROP_PATH_RATE: 0.3 PATCH_NORM: True PRETRAIN_IMG_SIZE: 384 WEIGHTS: "swin_large_patch4_window12_384_22k.pkl" PIXEL_MEAN: [123.675, 116.280, 103.530] PIXEL_STD: [58.395, 57.120, 57.375] ONE_FORMER: NUM_OBJECT_QUERIES: 250 INPUT: MIN_SIZE_TRAIN: !!python/object/apply:eval ["[int(x * 0.1 * 640) for x in range(5, 21)]"] MIN_SIZE_TRAIN_SAMPLING: "choice" MIN_SIZE_TEST: 640 MAX_SIZE_TRAIN: 2560 MAX_SIZE_TEST: 2560 CROP: ENABLED: True TYPE: "absolute" SIZE: (640, 640) SINGLE_CATEGORY_MAX_AREA: 1.0 COLOR_AUG_SSD: True SIZE_DIVISIBILITY: 640 # used in dataset mapper FORMAT: "RGB" TEST: DETECTIONS_PER_IMAGE: 250 EVAL_PERIOD: 5000 AUG: ENABLED: False MIN_SIZES: [320, 480, 640, 800, 960, 1120] MAX_SIZE: 4480 FLIP: True

I can tell you that there is nothing malicious in the files and so long as you change the permissions of the .yaml files being loaded to read only and make sure that you set the access so that only administrators can modify these files you can ensure that only these non-malicious versions stay on your machine.

TLDR:

I recommend setting it so that ONLY ADMINISTRATORS can edit/modify these files:

...\stable-diffusion-webui\extensions\sd-webui\controlnet\annotator\oneformer\detectron2\config\config.py

...\stable-diffusion-webui\extensions\sd-webui\controlnet\annotator\oneformer\detectron2\config\lazy.py

...\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\oneformer\configs\ade20k\Base-ADE20K-UnifiedSegmentation.yaml

...\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\oneformer\configs\ade20k\oneformer_swin_large_IN21k_384_bs16_160k.yaml

Then, the model should be safe to use since I do not see any unsafe code/loadings if these files are protected. Or you can just use the other models.

to my knowledge ip2p doesn't actually use a preprocessor. preprocessors are needed for adapters which expect the input image to look a certain way. they're filters, basically. ip2p doesn't need one because it uses a text prompt.

(pardon the necro i figure someone coming across this comment might like to know)

Really just personal preference, and it depends on the situation and what you're trying to achieve! Here's a guide on getting started with the basics, and what the most popular ones can do - https://education.civitai.com/civitai-guide-to-controlnet/

You want them all as each serve different purpose

chatgpt helps.

@darkthronefernandao265 it didn't 2 years ago when I asked the question. 😉

Civitai renames attached files based on Model Name and Version name, so yes, you could definitely rename the files once you've downloaded them to make it clearer.

Did you install the extension?
https://github.com/Mikubill/sd-webui-controlnet

I have this Issue too. Version 1.7.0. If someones knows how too fix it, please share.

Just use the All Control and select your preprocessor and model manually, and I see you've done that haha. I think that is the intended way to use it, the extension was intended to use with the vanilla controlnet.

Sorry for late response, you probably figured it by now, but I also had the same error. Just had to change the checkpoint to another model because the current one didn't work.

you dont

you need to hack into the mainframe

You get a hamster wheel for the additional processing power.

Try "Draw Things" app, it's in the app store. Runs locally.