Spaces:

---

Duplicated from  fancyfeast/joy-caption-alpha-two-vqa-test-one

fancyfeast

/

joy-caption-beta-one

like 647

Running on Zero

Same here - the space stopped working earlier today. Same for all other spaces by fancyfeast.

its not working, its giving me a runtime error

Thank you for the report. I restarted the space but it's still running into an issue. Might be a problem on HF's side so I'll try restarting it again later.

restart this like in next weekends in Friday and saturday's

Okay! Very thanks, I will try it :)

Seems to be working now

yes, it's working, thanks a lot!

It got an Runtime error and doesn't work. :/

Hi, it seems to have problems again. I get an error most of the time running it

Hi, it seems to have problems again. I get an error most of the time running it

Thank you for letting me know. Looks to be running okay right now?

tried right now and the first attempt was good, the second I got an error again

3 OK, 4 Error, 5 Error, 6 Error, 7 Ok, 8 Error and so on

Weird, I tried it a few times and no error. I'll give it a reboot later to see if that clears things up.

It seems, like its broken again...

I should mention that I use it via API

Any help would be appreciated. I use Joycaption constantly to caption single images but now I want to caption a dataset with 170 images. And it is totally unreliable.
The only thing i get is error and I don't know why, because with the GUI it works, for the same images.

I'm calling

https://fancyfeast-joy-caption-beta-one.hf.space/gradio_api/call/chat_joycaption

with this in body:

{
"data": [
{
"path": "{{352.webContentLink}}",
"meta": {
"_type": "gradio.FileData"
}
},
"Write a detailed description for this image in 260 words or less. If there is a person/character in the image you must refer to them as Cynthia. Describe only what is visible, do not invent new framing or elements. Always include information about lighting, camera angle, shot type (choose exactly one: extreme close-up, close-up, medium close-up, medium shot, cowboy shot, medium wide shot, wide shot, extreme wide shot), depth of field (background in focus or blurred), skin texture, hair detail, clothing fit and material, and background environment. Never mention her eye-color. Also include subjective aesthetic quality from low to very high, and describe the mood or atmosphere conveyed. State clearly whether the image is sfw, suggestive, or nsfw (choose exactly one). Use vulgar slang and profanity such as “fucking,” “slut,” “cock,” etc., do not use polite euphemisms. First describe the objective physical elements, then add subjective qualities. Do not explain or justify, only describe. Output as one coherent paragraph, no bullet points, no meta phrases.",
0.6,
0.9,
512,
false
]
}

This call returns 200 and a valid ID. But the GET action fails with:

https://fancyfeast-joy-caption-beta-one.hf.space/gradio_api/call/chat_joycaption/event_id

event: error
data: null

@nicras You can go over to https://captionator.ai, log in (free), and caption up to 500 images at a time (Joycaption backend). It will even let you download the captions in various formats, TXT, CSV, JSON.

@Ataylorm Thank you for the info and I tried but the page seems somewhat under construction and I need API functionality

@nicras , I am going to be launching the API in the next few days (Probably early next week), as well as ComfyUI nodes. If you are interested, you can be my first API guinea pig.

@Ataylorm Tell me, when its ready... I'll definitely give it a try

Found the failure...I used Authentication in header instead of Authorization. Now everything works like a charm.
As most of the time, the failure sits in front of the monitor.

The weird thing is, that it worked sometimes with Authentication but may be a session or token thing...

Thanks anyways for this great model!