Yes, I understand that. However the models Cloudflare is providing are capable of producing content

Yes, I understand that. However the models Cloudflare is providing are capable of producing content which should 100% not be possible with any model out there. I don't want to exactly discuss this in public, however my website got targeted by multiple prompts generating very very bad images.

SSomeyAI Yes, I understand that. However the models Cloudflare is providing are capable o...

Isaac McFadyen•1/14/25, 10:17 PM

Yes, but even then, Cloudflare cannot do anything about it. They are given the models by the people who train/produce them (Meta, independent authors, etc) and just host the models. They have no control over anything the model does/does not generate.

SomeyAIOP•1/14/25, 10:17 PM

Is there any way I can check images Cloudflare produces against prohibited contents?

SSomeyAI Is there any way I can check images Cloudflare produces against prohibited conte...

Isaac McFadyen•1/14/25, 10:19 PM

If it's related to CSAM then you can enable the CSAM scanning tool on your site https://developers.cloudflare.com/cache/reference/csam-scanning/ but that isn't realtime/instant. Other than that, there's models like LLaMa Guard (on Workers AI) for filtering text content but I don't believe there's a model on Workers AI for filtering image content.

Muhammadreza Haghiri•1/20/25, 8:19 PM

Hello. Just a question, it is still impossible to host our own models. Will it be possible anytime soon?

Original message was deleted

Isaac McFadyen•1/27/25, 4:34 PM

Try cloudflare-go, this channel is for Cloudflare's AI products.

Spotnag•1/27/25, 4:36 PM

did you have to delete it? i can't copy paste it there now and wil have to rewrite it... -.-

SSpotnag did you have to delete it? i can't copy paste it there now and wil have to rewri...

Isaac McFadyen•1/27/25, 4:39 PM

Oh whoops, sorry.

lfdepombo•1/27/25, 10:40 PM

Hi, I ran the example here and it works without issue - https://developers.cloudflare.com/workers-ai/models/llava-1.5-7b-hf/ I then used an image downloaded from the whatsapp api and I can download it correctly without a problem in my local machine or KV. However, the AI worker returns

  (error) Error processing message: AiError: 3010: Invalid or incomplete input for the model: model returned: Failed to decode image: cannot identify image file <_io.BytesIO object at 0x7ee1b5d57a10>

  (error) Error processing message: AiError: 3010: Invalid or incomplete input for the model: model returned: Failed to decode image: cannot identify image file <_io.BytesIO object at 0x7ee1b5d57a10>

Cloudflare Docs

llava-1.5-7b-hf · Cloudflare Workers AI docs

LLaVA is an open-source chatbot trained by fine-tuning LLaMA/Vicuna on GPT-generated multimodal instruction-following data. It is an auto-regressive language model, based on the transformer architecture.

lfdepombo•1/27/25, 10:46 PM

any idea why this would happen in the worker, but not locally?

Vero•1/28/25, 9:31 AM

------------------------------

This channel is archived, please use workers-ai ai-gateway vectorize

------------------------------

Yes, I understand that. However the models Cloudflare is providing are capable of producing content

Similar Threads