GitHub Community
QwenVL instruct abliterated node help with installing... #186432
Replies: 5 comments 44 replies
-
|
To install the Qwen-VL Instruct Abliterated node, first make sure you have Python 3.9 or newer installed on your system. Then install the required packages by running pip install torch transformers accelerate. After that, clone the repository that contains the node, go into the project folder, and install the dependencies using pip install -r requirements.txt. Next, download the Qwen-VL Instruct Abliterated model files from an official source like Hugging Face or GitHub and place them in the correct models folder. Finally, restart your application or server, and everything should be ready to use. |
Beta Was this translation helpful? Give feedback.
-
Thank you for the reply! So sorry for bothering you with this. I still can't get it to show in the dropdown or for that matter, get it to function as it should. (attached pic). What could I possibly be missing. As I said before, I'm am brand new to this, so obviously I probably need more direction than I thought. |
Beta Was this translation helpful? Give feedback.
-
|
This isn’t a Python or model issue, it’s a ComfyUI custom node registration issue. In ComfyUI, a node will never appear in the dropdown unless all three of these are true:
What to do step-by-step:
Also note:
So the absence from the dropdown usually means node load failure, not wrong model files. Fix the node load first, the model selection comes after. |
Beta Was this translation helpful? Give feedback.
-
|
What you pasted is only the IMPORT FAILED summary, not the actual error. The real blocker is the traceback lines immediately below:
Do this (Windows Portable, correct pip)Open terminal in the ComfyUI_windows_portable folder and run: Then close ComfyUI and run After restart, scroll a bit further down in the terminal and copy/paste:
|
Beta Was this translation helpful? Give feedback.
-
|
Sorry, more technical problems that I'm having. I figured to just take screenshots. I couldn't stay in the terminal to do that. Hope this is actually what you need. attached 2 pics |
Beta Was this translation helpful? Give feedback.
-
|
I just realized there is nothing in this window referring to the qwen file. Now I'm super confused. |
Beta Was this translation helpful? Give feedback.
-
|
Microsoft Windows [Version 10.0.26200.7623] C:\Users\Fyve\Desktop\ComfyUI_windows_portable>run_cpu.bat C:\Users\Fyve\Desktop\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --cpu --windows-standalone-build ComfyUI-Manager: installing dependencies done.** ComfyUI startup time: 2026-02-05 19:18:55.264 Prestartup times for custom nodes: Checkpoint files will always be loaded safely. The above exception was the direct cause of the following exception: Traceback (most recent call last): Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_audio.py module for custom nodes: Could not load this library: C:\Users\Fyve\Desktop\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torchaudio\lib\libtorchaudio.pyd The above exception was the direct cause of the following exception: Traceback (most recent call last): Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_lt_audio.py module for custom nodes: Could not load this library: C:\Users\Fyve\Desktop\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torchaudio\lib\libtorchaudio.pyd The above exception was the direct cause of the following exception: Traceback (most recent call last): Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\comfy_extras\nodes_audio_encoder.py module for custom nodes: Could not load this library: C:\Users\Fyve\Desktop\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torchaudio\lib\libtorchaudio.pyd Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-FramePackWrapper module for custom nodes: Torch not compiled with CUDA enabled Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-GGUF-FantasyTalking module for custom nodes: (unicode error) 'unicodeescape' codec can't decode bytes in position 149-150: truncated \UXXXXXXXX escape (nodes.py, line 608) Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Huihui-Qwen3-VL-4B-Instruct-abliterated module for custom nodes: [Errno 2] No such file or directory: 'C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Huihui-Qwen3-VL-4B-Instruct-abliterated\init.py' The above exception was the direct cause of the following exception: Traceback (most recent call last): Cannot import C:\Users\Fyve\Desktop\ComfyUI_windows_portable\ComfyUI\custom_nodes\comfyui-hunyuanvideowrapper module for custom nodes: Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback): Loading: ComfyUI-Impact-Pack (V8.28.2)[Impact Pack] Wildcard total size (0.00 MB) is within cache limit (50.00 MB). Using full cache mode. Loading: ComfyUI-Manager (V3.39.2)[ComfyUI-Manager] network_mode: public ComfyUI Revision: 428 [cb459573] *DETACHED | Released on '2026-02-05'[MultiGPU Core Patching] Patching mm.soft_empty_cache for Comprehensive Memory Management (VRAM + CPU + Store Pruning)
|
Beta Was this translation helpful? Give feedback.
-
|
You are running CPU Torch ( From your logs, the blocker is not “deprecation warnings”. You have real import failures, and two are directly relevant:
.\python_embeded\python.exe -m pip install -U "transformers>=4.47" accelerate sentencepiece einops torchvision timm pillowThen restart ComfyUI and re-check if |
Beta Was this translation helpful? Give feedback.
-
|
I spent the morning fixing other stuff that got broken yesterday. I lost python (I think on the last prompt that we did) and sage attention got broken also. I just fixed the sage and python this morning. The Qwen VL is totally broke now. I even started to do a brand new fresh install of Comfyui and wanted to just try and put that one workflow (SVI Pro) and as far as that goes, I'm totally lost and confused because almost everything is in the red, so I reverted back to the original again and that's where I'm at now. The SVI Pro workflow works but the Qwen is totally broken now (I have a cluster of files that I think I should just delete and start over with that). I'm not sure how to even back track and find out why because it seems I'm just going in circles with it. Bottom line is I'm not experienced enough to really know what goes where and so on. 3 weeks in and I think I'm doing okay but still have a ton to learn. I tried out some of the youtube videos but most of them are too vague for me since I'm just starting out with this. I will probably still want to pursue how in the heck I get this thing running like I want it to. My gosh! Sort of a nightmare the last 4 days. |
Beta Was this translation helpful? Give feedback.
-
|
By the way, do you think that it's a good idea for me to install cuda? You mentioned that alot of things depend on it. I think I can do that with no problem since I have an rtx 5090. |
Beta Was this translation helpful? Give feedback.
-
|
You don’t “install CUDA” to fix ComfyUI. On Windows + ComfyUI Portable, what matters is:
Right now your logs show If you have an RTX 5090: yes, you should run CUDA-PyTorch (not Toolkit).Do this from the ComfyUI_windows_portable folder:
You want Your QwenVL “workflow vs command line” confusionThat’s normal: the command-line chat can work while the ComfyUI node fails to import because the node depends on very specific Python packages and versions. Your actual QwenVL errors were:
So the right move is not “keep tinkering”. It’s reset only the Qwen nodes:
Key principle (to stop the “circles”)You currently have 10+ unrelated broken nodes. Fixing everything at once guarantees chaos. |
Beta Was this translation helpful? Give feedback.
-
|
'Could not find files for the given pattern(s)" |
Beta Was this translation helpful? Give feedback.
-
|
That error means Windows cannot execute the embedded Python binary. Until that’s fixed, don’t touch transformers/torch/etc. 1) Confirm the file exists (this is the correct check)In Command Prompt opened inside dir python_embeded\python.exe
2) If Windows says “This app can’t run on your PC”That usually means one of these:
3) Fastest fix (recommended): re-download + re-extract correctlyDo not try random repairs.
python_embeded\python.exe --versionIf that works, you’re back to a stable base. 4) If you still get blockedCheck Windows Security:
5) Only after Python runs: switch to CUDA PyTorch (since you have a 5090)When the embedded python works, then (and only then) do: python_embeded\python.exe -m pip uninstall -y torch torchvision torchaudio
python_embeded\python.exe -m pip install --upgrade torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu124
python_embeded\python.exe -c "import torch; print(torch.__version__, torch.cuda.is_available())"You want |
Beta Was this translation helpful? Give feedback.
-
|
Alot of things broke yesterday when we did that prompt to find the node (I can't remember the prompt). Anyhow, as of now, I might have a solution to the errors we're getting now or not. I have comfyui inside another folder at the moment because I was obviously juggling things this morning, so let me take the folder out of there and see if this is the case. Right now I need to restart because something is open and it won't let me move the folder out. |
Beta Was this translation helpful? Give feedback.
-
|
Sorry, you lost me again on NUmber2. I have the new comfy Portable from this morning and that one opens up. Should I use the python embeded file from that and put it in my portable that was originally running? |
Beta Was this translation helpful? Give feedback.
-
|
Do NOT mix folders between installs. Copying Use this rule: If the new portable runs
|
Beta Was this translation helpful? Give feedback.
-
|
Okay, yes, that was what I was doing this morning and I got fed up because the whole SVL workflow had red nodes all over. Anyhow, so far, I installed a couple of things on the new one. I updated the python to 3.13.11, which is showing now and I reinstalled sage attention and I put the workflow of SVI pro on there because that is really the only one I use now. This will give me a chance to free up a ton of memory by deleting the files I am not going to use. Okay, sorry, getting back to this and yes we can move forward since this isn't corrupt any longer. |
Beta Was this translation helpful? Give feedback.
-
|
Okay, now I got everything running except for the QwenVL and I also noticed the 'Install Missing Nodes' is finally working. I think it was corrupted for a couple of weeks. Anyhow, I will stop at this point and wait for your guidance on how to get this generator to function. Thanks for all of the help today! |
Beta Was this translation helpful? Give feedback.
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
-
Body
Hello, I am new to Comfy. I hope I am posting in the correct place. I was wondering if I could possibly get help with the full installation of Qwen-vl-instruct abliterated. I've tried what I thought, everything possible, to get this to function and I don't know what I am doing wrong (the ablit.version doesn't even show up in the node dropdown). Qwen instruct does function on my computer in Comfyui (in an SVI Pro workflow), but I cannot seem to get the correct setup for running the abliterated version. I would apprectiate the help with a clear guide, if possible and if I am in the wrong area for questions and answers, please redirect me. Thank you very much!
Guidelines
Beta Was this translation helpful? Give feedback.
All reactions