A relatively simple LoRA testing workflow that can be used to compare multiple LoRAs (or one LoRA at difference weights) with less clicks. I cleaned it up a bit and added some notes for easier use.
V2 - Added a group for enabling or disabling Lora's. The wires are a bit messy on this version compared to V1
Brief Instructions for using: PLEASE CHANGE YOUR WIRE SETTINGS TO 'STRAIGHT' FOR OPTIMAL WIRE MANAGEMENT.
-First thing you need to do is go to your settings via ⚙️icon
-Next scroll down the list of options (you may have many if you have custom nodes installed) and find the setting called “Link Render Mode”.
-By default this would be set to Spline, but you can change it 'straight' for optimal wire management in this workflow.
Instructions with images:
https://weirdwonderfulai.art/comfyui/tip-workflow-link-noodle-settings/
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
- All Instruction nodes are bright yellow for easy locating.
-To install missing nodes, open the Manager and select "install Missing Custom Nodes"
-To turn off a group, Right click on the group and select "Set Group Nodes To Never" or "Bypass Group nodes"
-To turn on a group, right click on the group and select "Set Group Nodes To Always"
- Although this workflow is relatively straight forward; if you have further questions, please do not hesitate to leave a comment.
For adjusting the amount of LoRA's you want to use on V1, mute and bypass accordingly: 
I prefer V1 due to it's cleanliness, but to each their own.
Description
FAQ
Comments (6)
you're my hero
Nice job on this, it works well. I updated for Flux and have been testing it.
One thing I noticed is that if I don't have 8 loras I want to test, I can't just bypass the lora loader or the workflow gives an error (I think because the label section then cannot pull in a lora name for those items). Have you run into this issue? Do you just also bypass the corresponding labelers? It gets a bit messy when doing so.
Hmm I didn't think of this when making it because I usually test it for 8 epochs when I train a model... My best suggestion would be to just break up the top group to bypass the corresponding label. You could also potentially add switches if you can wire it properly. Let me open the work flow and take a quick look. If I get a simple enough solution I'll post a V2
(disclaimer, this comment is a lot shorter than the one I originally typed.. I just spent like 20 minutes typing a detailed one and refreshed my page.... RIP)
So i just tested it and it should not be erroring out. You will want to mute (not bypass) the Lora nodes on the left and it should run just fine. The ksamplers will turn red but the workflow wont stop. The downside to this is that the imagine combine does not work anymore. If you still want to combine the images, you will just want to bypass (not mute) the labelers and combiners for each LoRA.
To quickly do this, you can utilize the following shortcuts.
-Hold CTRL, click and drag your mouse to highlight the desired nodes.
-Once highlighted, use CTRL+M to mute the nodes.
-Alternatively you can use CTRL+B to bypass the nodes.
I am not going to add switches and republish for a few reasons.
1) the workflow should not be erroring out if you mute the LoRA nodes, and the image combine still functions with a simple bypass.
2) The fix listed above using the shortcuts takes about 15 seconds to adjust the workflow for a differnt number of LoRAs.
3) The required amount of wires and routing for the required switches would confuse the average user more than it would help them unless i redid the whole workflow, which would take a fair amount of time/work and ins't quite necessary due to the two reasons above. (I could spaghetti it, but I am not willing to post a spaghetti workflow.)
If/When I update this workflow however, I will take this into consideration and will likely add switches so it's both digestible for people, and functional. I added an image to the bottom of the description showing what it looks like using three LoRAs.
Thank you for your feedback! :)
Edit: I posted a new version with the switches LMAO. I still prefer V1 but you may like it (wire management is not as good as V1)
@TwinklingSquid I am sorry I sent you down a rabbit hole :). I very much appreciate the extra information and investigation, and the new version! I will give that a go ASAP. Thanks very much!
@EnragedAntelope Not all! I live for the rabbit holes. smirks