Acly
local server instalation
hello, when i start ai diffusion tool installation, i get this error. im sending multiple screenshots, it doesnt fit on one.
client.log
server.log
Hm, it's quite a lot of text to tell you that there's an issue with the internet connection, and the download timed out.
Probably because previous installation was aborted and some files are leftover. Delete the folder to retry in the same location.
Having a similar problem after making some efficiency changes to the launch script, I didn't change anything else as far as I know. It's telling me it can't find the CLIP Vision module, but It's definitely in there. Unless it actually needs ALL the ones on this message? But it worked fine when I first installed it.
Edit: Turns out it was in a higher folder for some dumb reason, but now it's telling me it's missing ip-adapter_sd15, but that is in there, and it's not in the higher folder.
After much digging and searching and cursing I have discovered a possible explanation for this problem, though not a solution as yet.
After installing comfy via this addon, it worked in krita and I made some test inpainting things. That was fun. Then, because I want to use ComfyUI for more than just krita, I accessed it directly, installed the manager, got the node packs I like, and after all that I tested it with krita again, and it still worked.
Cut to the next day, and it no longer works.
So I had a read through the logs in Krita from trying to launch it via the config menu, and there are a lot of modules it's failing to load for some reason. After searching around I found somebody with a similar problem in a different program, where it was suggested that it's trying to load the wrong python instance. If that's happening here, then it would be loading the system python install instead of the one inside the Comfy folder, and of course it wouldn't find what it was looking for.
And of course it still won't connect to the server when I start it separately and try to get in via Custom Server, saying it can't find the sd15 files. The script I use to start the server does use the system python, so it's possible that the krita addon is still tracking from that starting point which could result in it looking for things in the wrong places.
So, with that in mind, what is the solution? Is there any way to tell the krita addon to look somewhere specific?
The managed install is an easy way to get things working if you don't need customization.
If you want a custom ComfyUI (with manager, new nodes, etc.) install it separately and set it up according its installation instructions and https://github.com/Acly/krita-ai-diffusion/wiki/ComfyUI-Setup - run it externally and connect via Custom Server.
Trying some hybrid mix of the two (sounds like that's what you're doing?) is just going to result in pain.
That's what I tried the first time, and I had the same problem, it doesn't see the sd15 files.
For example, here's the client log from an attempt to load it a few minutes ago:
2024-12-14 13:52:54,301 INFO Connecting to http://127.0.0.1:7860 2024-12-14 13:52:54,616 INFO Optional Text Encoder model clip_l for All not found (search path: clip_l) 2024-12-14 13:52:54,616 INFO Optional Text Encoder model clip_g for All not found (search path: clip_g) 2024-12-14 13:52:54,616 INFO Optional Text Encoder model t5 for All not found (search path: t5) 2024-12-14 13:52:54,616 INFO Optional Text Encoder model clip_l for All not found (search path: clip_l) 2024-12-14 13:52:54,616 INFO Optional Text Encoder model clip_g for All not found (search path: clip_g) 2024-12-14 13:52:54,616 INFO Optional Text Encoder model t5 for All not found (search path: t5) 2024-12-14 13:52:54,616 INFO Found Image Encoder (VAE) model default for SD 1.5: vae-ft-mse-840000-ema-pruned.ckpt 2024-12-14 13:52:54,616 INFO Found Image Encoder (VAE) model default for SD XL: sdxl_vae.safetensors 2024-12-14 13:52:54,617 INFO Found Image Encoder (VAE) model default for Flux: sdxlVAE_sdxlVAE.safetensors 2024-12-14 13:52:54,617 WARNING Missing ControlNet model inpaint for SD 1.5 2024-12-14 13:52:54,617 INFO -> No model matches search paths: control_v11p_sd15_inpaint 2024-12-14 13:52:54,617 INFO -> Available models: depthnsfwanimeposes_v60/l_ly/lora/fucsil1-000011.safetensors, depthnsfwanimeposes_v60/l_ly/lora/viewer on leash 6.safetensors, depthnsfwanimeposes_v60/l_ly/lycoris/saya-hand in panties.safetensors, controlnet852a_normal.safetensors, diffusion_pytorch_model_promax.safetensors, xinsir-controlnet-union-sdxl-1.0-promax.safetensors 2024-12-14 13:52:54,617 INFO Optional ControlNet model inpaint for Flux not found (search path: flux.1-dev-controlnet-inpaint) 2024-12-14 13:52:54,617 INFO Found ControlNet model universal for SD XL: xinsir-controlnet-union-sdxl-1.0-promax.safetensors 2024-12-14 13:52:54,617 INFO Optional ControlNet model scribble for SD 1.5 not found (search path: control_v11p_sd15_scribble, control_lora_rank128_v11p_sd15_scribble) 2024-12-14 13:52:54,617 INFO Optional ControlNet model scribble for SD XL not found (search path: xinsirscribble, scribble-sdxl, mistoline_fp16, mistoline_rank, control-lora-sketch-rank, sai_xl_sketch_) 2024-12-14 13:52:54,617 INFO Optional ControlNet model line_art for SD 1.5 not found (search path: control_v11p_sd15_lineart, control_lora_rank128_v11p_sd15_lineart) 2024-12-14 13:52:54,617 INFO Optional ControlNet model line_art for SD XL not found (search path: xinsirscribble, mistoline_fp16, mistoline_rank, scribble-sdxl, control-lora-sketch-rank, sai_xl_sketch_) 2024-12-14 13:52:54,617 INFO Optional ControlNet model line_art for Flux not found (search path: mistoline_flux) 2024-12-14 13:52:54,617 INFO Optional ControlNet model soft_edge for SD 1.5 not found (search path: control_v11p_sd15_softedge, control_lora_rank128_v11p_sd15_softedge) 2024-12-14 13:52:54,617 INFO Optional ControlNet model soft_edge for SD XL not found (search path: mistoline_fp16, mistoline_rank, xinsirscribble, scribble-sdxl) 2024-12-14 13:52:54,617 INFO Optional ControlNet model soft_edge for Flux not found (search path: mistoline_flux) 2024-12-14 13:52:54,617 INFO Optional ControlNet model canny_edge for SD 1.5 not found (search path: control_v11p_sd15_canny, control_lora_rank128_v11p_sd15_canny) 2024-12-14 13:52:54,617 INFO Optional ControlNet model canny_edge for SD XL not found (search path: xinsircanny, canny-sdxlcontrol-lora-canny-rank, sai_xl_canny_) 2024-12-14 13:52:54,617 INFO Optional ControlNet model canny_edge for Flux not found (search path: flux-canny, mistoline_flux) 2024-12-14 13:52:54,617 INFO Optional ControlNet model depth for SD 1.5 not found (search path: control_sd15_depth_anything, control_v11f1p_sd15_depth, control_lora_rank128_v11f1p_sd15_depth) 2024-12-14 13:52:54,617 INFO Optional ControlNet model depth for SD XL not found (search path: xinsirdepth, depth-sdxl, control-lora-depth-rank, sai_xl_depth_) 2024-12-14 13:52:54,617 INFO Optional ControlNet model depth for Flux not found (search path: flux-depth) 2024-12-14 13:52:54,617 INFO Optional ControlNet model normal for SD 1.5 not found (search path: control_v11p_sd15_normalbae, control_lora_rank128_v11p_sd15_normalbae) 2024-12-14 13:52:54,617 INFO Optional ControlNet model pose for SD 1.5 not found (search path: control_v11p_sd15_openpose, control_lora_rank128_v11p_sd15_openpose) 2024-12-14 13:52:54,617 INFO Optional ControlNet model pose for SD XL not found (search path: xinsiropenpose, openpose-sdxl, control-lora-openposexl2-rank, thibaud_xl_openpose) 2024-12-14 13:52:54,617 INFO Optional ControlNet model segmentation for SD 1.5 not found (search path: control_v11p_sd15_seg, control_lora_rank128_v11p_sd15_seg) 2024-12-14 13:52:54,617 INFO Optional ControlNet model segmentation for SD XL not found (search path: sdxl_segmentation_ade20k_controlnet) 2024-12-14 13:52:54,617 WARNING Missing ControlNet model blur for SD 1.5 2024-12-14 13:52:54,617 INFO -> No model matches search paths: control_v11f1e_sd15_tile, control_lora_rank128_v11f1e_sd15_tile 2024-12-14 13:52:54,617 INFO -> Available models: depthnsfwanimeposes_v60/l_ly/lora/fucsil1-000011.safetensors, depthnsfwanimeposes_v60/l_ly/lora/viewer on leash 6.safetensors, depthnsfwanimeposes_v60/l_ly/lycoris/saya-hand in panties.safetensors, controlnet852a_normal.safetensors, diffusion_pytorch_model_promax.safetensors, xinsir-controlnet-union-sdxl-1.0-promax.safetensors 2024-12-14 13:52:54,617 INFO Optional ControlNet model blur for SD XL not found (search path: xinsirtile, tile-sdxl, ttplanetsdxlcontrolnet, ttplanet_sdxl_controlnet_tile_realistic, ttplanet_controlnet_tile_realistic) 2024-12-14 13:52:54,618 INFO Optional ControlNet model blur for Flux not found (search path: flux.1-dev-controlnet-upscale) 2024-12-14 13:52:54,618 INFO Optional ControlNet model stencil for SD 1.5 not found (search path: control_v1p_sd15_qrcode_monster) 2024-12-14 13:52:54,618 INFO Optional ControlNet model stencil for SD XL not found (search path: sdxl_qrcode_monster) 2024-12-14 13:52:54,618 INFO Optional ControlNet model hands for SD 1.5 not found (search path: control_sd15_inpaint_depth_hand) 2024-12-14 13:52:54,618 INFO Optional ControlNet model hands for SD XL not found (search path: control-lora-depth-rank, sai_xl_depth_) 2024-12-14 13:52:54,618 WARNING Missing CLIP Vision model ip_adapter for All 2024-12-14 13:52:54,618 INFO -> No model matches search paths: sd1.5/pytorch_model.bin, sd1.5/model.safetensors, clip-vision_vit-h.safetensors, clip-vit-h-14-laion2b-s32b-b79k 2024-12-14 13:52:54,618 INFO -> Available models:
There is no way it should be able to see the nsfw pose things, but not anything else it's looking for, because they are all there from when they were installed by the addon, a few days ago. I don't even remember when I got those but they are in the A1111 file tree that I have comfy set to look into for resources, which also did work when I tried it the first day of this install. They are in /home/USERNAME/Stable/A1111-new/stable-diffusion-webui/models/ControlNet with a few other things that it's not seeing either, but they're not core anyway so I wouldn't expect it to see them.
None of this makes sense to me.
Oh yeah and I forgot, the server puts out an error when krita tries to connect to it:
fatal: not a git repository (or any of the parent directories): .git Failed to get ComfyUI version: Command '['git', 'describe', '--tags']' returned non-zero exit status 128.
No idea what that's about but it may be pertinent.
Oh yeah and I forgot, the server puts out an error when krita tries to connect to it: fatal: not a git repository (or any of the parent directories): .git Failed to get ComfyUI version: Command '['git', 'describe', '--tags']' returned non-zero exit status 128.
That can be ignored, it's only used to print its own version.
Regarding missing files, the plugin doesn't look for any files directly. It only sees what ComfyUI reports, and that's what it's in its models folder and the extra paths you configured. If you create a Load ControlNet node in Comfy web UI and it shows the file, the plugin should work. If it doesn't, you messed up the paths somewhere.
What do you mean it works, does it actually show the models it's searching for? Or just some models?
Does eg. Load CLIP Vision node show clip-vision_vit-h.safetensors or one of the other names mentioned in the log?
Ok, clip vision shows nothing, that's progress. So what folder is that supposed to be in? And how did it work the first time if it's not in there now?
Ok, I figured it out!
In extra_model_paths.yaml I had to enable the comfyui section and add a few lines:
comfyui: base_path: /home/USERNAME/Stable/ComfyUI # You can use is_default to mark that these folders should be listed first, and used as the default dirs for eg downloads is_default: true checkpoints: models/checkpoints/ clip: models/clip/ clip_vision: models/clip_vision/ configs: models/configs/ controlnet: models/controlnet/ diffusion_models: | models/diffusion_models models/unet embeddings: models/embeddings/ inpaint: models/inpaint ipadapter: models/ipadapter/ loras: models/loras/ upscale_models: models/upscale_models/ vae: models/vae/
is there no way to make it format that code block properly?