What happened to the quality of the `dreamshaper-8-lcm` model? Its outputs have been of extremely lo

What happened to the quality of the

dreamshaper-8-lcm

dreamshaper-8-lcm

dreamshaper-8-lcm

dreamshaper-8-lcm model? Its outputs have been of extremely low quality for the past ~3 days, to the point that they are completely unusable. Below are two examples, 1 month apart. Same prompt.

TThiago What happened to the quality of the `dreamshaper-8-lcm` model? Its outputs have ...

Raylight•7/14/24, 8:56 PM

Looks like the "guidance" parameter has been exposed/activated and it's possible that it doesn't behave as intended. The default of 7.5 will give results like the image on the right. If you add "guidance: 0.0" (or maybe 1.0) you'll get the same or similar results as you got before. You'll get better prompt following with higher numbers, but as of now there's a narrow range before images become degraded to the point of being unusable.

RRaylight Looks like the "guidance" parameter has been exposed/activated and it's possible...

ThiagoOP•7/14/24, 9:53 PM

Oh, that makes perfect sense. Thank you!

akazwz•7/15/24, 12:00 PM

@Rahul it need to use vision model like LLaVA https://developers.cloudflare.com/workers-ai/models/#image-to-text

Cloudflare Docs

Models · Cloudflare Workers AI docs

Browse our entire catalog of models:

akazwz•7/15/24, 12:03 PM

if you want to use the image url in the prompt. you can use function calling to create a tools that use to be vision the image

Aakazwz Hello, I build a LLMs/AI chat platform build and design for Cloudflare worker AI...

akazwz•7/15/24, 12:06 PM

@Rahul In this code. i can simple answer with the image. and u can put it in tools https://github.com/akazwz/antonai/blob/main/app/routes/api.chat.ts#L70

GitHub

antonai/app/routes/api.chat.ts at main · akazwz/antonai

a LLMs/AI chat platform build and design for Cloudflare worker AI. with many features like function calling, image to text, text to image, text to speech - akazwz/antonai

akazwz•7/15/24, 12:21 PM

i write a simple tools to make this work.

akazwz•7/15/24, 1:32 PM

maybe you need some tools to parse the pdf like langchain. i am not sure if it will work

RRahul Thanks <@959078620360486964>

akazwz•7/16/24, 6:19 AM

You're welcome

Ken•7/18/24, 9:45 AM

Ken•7/18/24, 9:45 AM

server down

Ken•7/18/24, 9:45 AM

Ken•7/18/24, 9:45 AM

@cb

KKen Click to see attachment

cb•7/18/24, 10:04 AM

Working for me right now.

Ken•7/18/24, 10:08 AM

naa bro

Ken•7/18/24, 10:08 AM

its not

Ken•7/18/24, 10:08 AM

@cb

Ken•7/18/24, 10:08 AM

KKen <@462316658082709514>

DaniFoldi•7/18/24, 10:12 AM

?pings

Flare•7/18/24, 10:12 AM

Please do not ping community members for non-moderation reasons. Doing so will not solve your issue faster and will make people less likely to want to help you.

threekat•7/19/24, 6:21 PM

poor lonlos has HIV

DDeleted User `@cf/runwayml/stable-diffusion-v1-5-img2img` is generating very low quality and ...

Raylight•7/19/24, 8:54 PM

Try adding the strength parameter. strength: 0.5strength: 0.5 for example. It defaults to 1.0, which pretty much means ignore the input image. Also, you could try to describe the desired output instead, like "a lion sitting on a bench". As for quality, the model - or something in Cloudflare's implementation - seems to have a tendency to produce over-saturated colors. Not sure what's going on. Setting the guidance parameter to a low value helps, but it's not really a solution.

DDeleted User how to make consistent characters from cf stable diffusion?

Isaac McFadyen•7/21/24, 1:27 PM

Stable Diffusion currently struggles with text, particularely the earlier variants like 1.5 and lower-step versions like lightning.

Isaac McFadyen•7/21/24, 1:27 PM

If you want to make readable text then Stable Diffusion 3 is your best bet, but I don't believe it's currently available on Workers AI.

reality•7/21/24, 3:01 PM

side by side of multi language probing for purpose based testing of multiple examples

broooo•7/22/24, 1:44 AM

like why not giving cloudflare vector for free?

0xCC•7/22/24, 6:02 AM

Hey. I found my model for text generation has reached limit - when the limit would refresh?

sam•7/22/24, 6:34 AM

Just to confirm, if I use a text generation beta model (@cf/meta/llama-3-8b-instruct) it will be completely free right?

Erisa•7/22/24, 8:53 AM

?channel-crossposting

Flare•7/22/24, 8:53 AM

Please do not post your question in multiple channels per the rules at No Access. It creates confusion for people trying to help you and doesn't get your issue or question solved any faster.

broooo•7/22/24, 12:38 PM

I am getting, You may need to enable the cache_option_enabled compatability flag how to resolve it?

broooo•7/22/24, 12:47 PM

TypeError: Unsupported cache mode: no-store