showing posts tagged #aihorde — clear

0

The AI Horde has a new Interface, a new Image generation frontend, new backend, and all new documentation!

Haidra An image generation studio and documentation on aihorde.net haidra.net
preview

We haven’t been posting a lot in this comm lately, but we’ve been still hard at work. I wanted to announce our latest release which includes an all-new interface for generating images via the AI Horde, this time hosted and developed by us directly: aihorde.net/generate/image

There’s a ton of improvement on the backend as we support all newer models (Anima, Krea2 etc) and we’re still complete free after 4 years of service.

If you don’t have access to a powerful GPU, feel free to use the AI Horde. And if you have any idle GPU, no matter it’s power, feel free to become a volunteer as well!

We haven’t been posting a lot in this comm lately, but we’ve been still hard at work. I wanted to announce our latest release which includes an all-new interface for generating images via the AI Horde, this time hosted and developed by us directly: aihorde.net/generate/image

There’s a ton of improvement on the backend as we support all newer models (Anima, Krea2 etc) and we’re still complete free after 4 years of service.

If you don’t have access to a powerful GPU, feel free to use the AI Horde. And if you have any idle GPU, no matter it’s power, feel free to become a volunteer as well!

0

Community-driven efficient LLM development - Possible?

preview

Hello everyone!

TL;DR: I want to propose a community-driven effort to research and improve 1-bit LLM models, for use within and outside the Horde. I think having access to such models would be very useful for the overall project, as you do not need a lot of compute to run “bigger” models if they’re compressed well. Relevant paper.

We currently live in very interesting times regarding AI development. The big companies in the US seem to still be ahead, but groups from China, working on open-weights models, are making very impressive strides. However, the focus has been and still is on developing really big models. Most of the impressive models that keep coming out are bigger and bigger, leaving most people to pay for API tokens if they want something useful. Probably this is also incentivised, as Nvidia wants to make more money from sales to data centers.

I have been keeping an eye out for the AI Horde project, and I always wanted to help out, but never got around to setting it properly up due to my AMD GPU and running Windows. However, really cool project, and I wanted to congratulate everyone involved!

In my opinion, the goal of the AI Horde community also positions it as one of the very few that are capable of bringing forward some LLMs that can be “smart” on consumer-level hardware. That is, getting models that can handle long-horizon tasks well without paying providers for API access.

Since the first paper on 1.68-bit LLMs, there has been quite a few other suggestions thrown around (recent example). No one really uses these models much, as the quality is seriously degraded. Similarly, no one is really trying to improve these models further, as there is no incentive to do so when you can just pay someone 3$ per million token output from an existing open-weight model. For example, to the best of my knowledge, no one has tried to introduce latent reasoning (example) in this context, or specifically training/fine-tuning models at 1-bit levels.

So, to get to the point, would it make sense to get some community-driven research in this area? I believe that we could all pool together compute, good training data, ideas for fine-tuning / RL-training, etc. If it works out, we could have a method that makes existing larger models (say, up to 200B) available on a single 24GB GPU.

First thing I would try is to expand on the recent NanoQuant paper:

  1. Wait for weights to be released.
  2. If no weights come out, quantize a Qwen 3 32B, and try a more diverse dataset, with more tokens to see if fidelity can improve. I could get some access to GPUs for this myself. Another option for 1-bit models would be using other existing ones (e.g., Unsloth), but performance degradation is much bigger in those versions, from what I have seen. Furthermore, the compression of these models is not as efficient, and you would not be able to fully run a 70B parameter (as described in the paper), with only 8GB VRAM.
  3. Get an LLM (or human volunteers) to determine behaviour on various tasks: find limitations, strengths, etc. Get some human preferences for RLHF, or use a bigger LLM to grade output quality. Preference here on logic tasks.
  4. Perform fine-tuning of 1-bit model based on the gathered data, and deploy for use. Return to step 2 after a while.

Fine-tuning the 1-bit model might get a bit hairy, as the binary operation are not differentiable. We also wouldn’t be able to up-cast back to F32 for regular training, as this would completely invalidate the consumer-driven access to these models. Simplest idea would be to train a LoRA head, or do some stochastic-driven training (e.g., flipping bits). However, the latter would probably be very unstable, and not work out, and LoRA might be the only option. I’m not a mathematician, so I am open to suggestions here :)

Past the initial prototype, I would consider the following stuff that’s already implemented in other quants:

  • Try to re-calibrate the LayerNorm / RMSNorm parameters based on the original model’s activations.
  • Perform some regular KL-divergence distillation
  • Other than using an LLM as a judge for RL, perhaps one could also fine-tune using a semantic similarity metric while aligning with the output of the original model. This could ensure that the intent is the same, even if the style differs.
  • Depending on token complexity, look into reducing compression just for difficult tokens and compressing further for easy ones (à la FlexQuant)
  • Mixture-of-Experts-style quantization, where we increase compression for experts that are not important, and reduce it for higher-frequency ones.

Curious what everyone thinks!

Hello everyone!

TL;DR: I want to propose a community-driven effort to research and improve 1-bit LLM models, for use within and outside the Horde. I think having access to such models would be very useful for the overall project, as you do not need a lot of compute to run “bigger” models if they’re compressed well. Relevant paper.

We currently live in very interesting times regarding AI development. The big companies in the US seem to still be ahead, but groups from China, working on open-weights models, are making very impressive strides. However, the focus has been and still is on developing really big models. Most of the impressive models that keep coming out are bigger and bigger, leaving most people to pay for API tokens if they want something useful. Probably this is also incentivised, as Nvidia wants to make more money from sales to data centers.

I have been keeping an eye out for the AI Horde project, and I always wanted to help out, but never got around to setting it properly up due to my AMD GPU and running Windows. However, really cool project, and I wanted to congratulate everyone involved!

In my opinion, the goal of the AI Horde community also positions it as one of the very few that are capable of bringing forward some LLMs that can be “smart” on consumer-level hardware. That is, getting models that can handle long-horizon tasks well without paying providers for API access.

Since the first paper on 1.68-bit LLMs, there has been quite a few other suggestions thrown around (recent example). No one really uses these models much, as the quality is seriously degraded. Similarly, no one is really trying to improve these models further, as there is no incentive to do so when you can just pay someone 3$ per million token output from an existing open-weight model. For example, to the best of my knowledge, no one has tried to introduce latent reasoning (example) in this context, or specifically training/fine-tuning models at 1-bit levels.

So, to get to the point, would it make sense to get some community-driven research in this area? I believe that we could all pool together compute, good training data, ideas for fine-tuning / RL-training, etc. If it works out, we could have a method that makes existing larger models (say, up to 200B) available on a single 24GB GPU.

First thing I would try is to expand on the recent NanoQuant paper:

  1. Wait for weights to be released.
  2. If no weights come out, quantize a Qwen 3 32B, and try a more diverse dataset, with more tokens to see if fidelity can improve. I could get some access to GPUs for this myself. Another option for 1-bit models would be using other existing ones (e.g., Unsloth), but performance degradation is much bigger in those versions, from what I have seen. Furthermore, the compression of these models is not as efficient, and you would not be able to fully run a 70B parameter (as described in the paper), with only 8GB VRAM.
  3. Get an LLM (or human volunteers) to determine behaviour on various tasks: find limitations, strengths, etc. Get some human preferences for RLHF, or use a bigger LLM to grade output quality. Preference here on logic tasks.
  4. Perform fine-tuning of 1-bit model based on the gathered data, and deploy for use. Return to step 2 after a while.

Fine-tuning the 1-bit model might get a bit hairy, as the binary operation are not differentiable. We also wouldn’t be able to up-cast back to F32 for regular training, as this would completely invalidate the consumer-driven access to these models. Simplest idea would be to train a LoRA head, or do some stochastic-driven training (e.g., flipping bits). However, the latter would probably be very unstable, and not work out, and LoRA might be the only option. I’m not a mathematician, so I am open to suggestions here :)

Past the initial prototype, I would consider the following stuff that’s already implemented in other quants:

  • Try to re-calibrate the LayerNorm / RMSNorm parameters based on the original model’s activations.
  • Perform some regular KL-divergence distillation
  • Other than using an LLM as a judge for RL, perhaps one could also fine-tune using a semantic similarity metric while aligning with the output of the original model. This could ensure that the intent is the same, even if the style differs.
  • Depending on token complexity, look into reducing compression just for difficult tokens and compressing further for easy ones (à la FlexQuant)
  • Mixture-of-Experts-style quantization, where we increase compression for experts that are not important, and reduce it for higher-frequency ones.

Curious what everyone thinks!

0

How to submit aesthetic ratings through v2 API?

preview

Hey!

Not sure if this is the palce to ask questions, but I have one.

I can’t quite figure out how to submit ratings for generated images.

According to the API docs, the /v2/generate/rate/{id} endpoint is for submitting ratings, but I can’t figure out how to actually call this enpoint.

For starters, what is {id} in this case?

I’ve checked the source at github, and to my understanding it’s a WaitingPrompt id, which, if I’m correct is the id what /api/v2/generate/async gives back in its response.

Now the API docs tells me I can submit a set of ratings and also a best one from that set. Is this means that if I generate multiple images with one request, I can rate only those images? And should I use the UUID of the images in each rating object, the id I get back from /v2/generate/status/{id} in each generation object?

Also, should I specify the apikey in headers?

I tried to send POST requests to https://aihorde.net/api/v2/generate/rate/%7Bid%7D, but I always get back the following message:

{'message':'The browser (or proxy) sent a request that this server could not understand.'}

In the request URL, I tried to use all the IDs I could think of. Job IDs, image IDs… I also realized in the meantime that ArtBot too generates UUIDs for itself…

My “final” request looked like this:

  • in the URL, I used the job id of my image generation request
  • In the message, I have put one rating, for which ID I used the id what I got in /v2/generate/status/{id} response. (generations[0].id)

As for why I try to send in ratings manually; the rating feature is broken in ArtBot (CORS errors, in all browser). As well as everywhere else.

I see references to the v1 API in websites that can in theory send in ratings to Horde, as well in the AI-Horde source code.

So, I’m kinda lost and confused with this rating thing. Maybe I’m doing things entirely wrong.

But anyway. Any help or answer is appreciated! :)

Thanks!

Hey!

Not sure if this is the palce to ask questions, but I have one.

I can’t quite figure out how to submit ratings for generated images.

According to the API docs, the /v2/generate/rate/{id} endpoint is for submitting ratings, but I can’t figure out how to actually call this enpoint.

For starters, what is {id} in this case?

I’ve checked the source at github, and to my understanding it’s a WaitingPrompt id, which, if I’m correct is the id what /api/v2/generate/async gives back in its response.

Now the API docs tells me I can submit a set of ratings and also a best one from that set. Is this means that if I generate multiple images with one request, I can rate only those images? And should I use the UUID of the images in each rating object, the id I get back from /v2/generate/status/{id} in each generation object?

Also, should I specify the apikey in headers?

I tried to send POST requests to https://aihorde.net/api/v2/generate/rate/%7Bid%7D, but I always get back the following message:

{'message':'The browser (or proxy) sent a request that this server could not understand.'}

In the request URL, I tried to use all the IDs I could think of. Job IDs, image IDs… I also realized in the meantime that ArtBot too generates UUIDs for itself…

My “final” request looked like this:

  • in the URL, I used the job id of my image generation request
  • In the message, I have put one rating, for which ID I used the id what I got in /v2/generate/status/{id} response. (generations[0].id)

As for why I try to send in ratings manually; the rating feature is broken in ArtBot (CORS errors, in all browser). As well as everywhere else.

I see references to the v1 API in websites that can in theory send in ratings to Horde, as well in the AI-Horde source code.

So, I’m kinda lost and confused with this rating thing. Maybe I’m doing things entirely wrong.

But anyway. Any help or answer is appreciated! :)

Thanks!

0

Haidra Canvas 2024 entry!

preview

Go to this URL and help us make a presence for Haidra and by extension /0 as well!

If you’re wondering what canvas 2024 is, check here.

If you’re wondering how this fits the instance theme, I invite you to check out this critical reading, which is what inspired the name

The template position is on the bottom right

You will need to login. Once you do, lower the opacity of the template, and place some pixels where there’s none

EDIT: Changed position to bottom right as it feels more natural to grow from the bottom up. I also increased its size/fidelity a bit since we had a bit more space there

Go to this URL and help us make a presence for Haidra and by extension /0 as well!

If you’re wondering what canvas 2024 is, check here.

If you’re wondering how this fits the instance theme, I invite you to check out this critical reading, which is what inspired the name

The template position is on the bottom right

You will need to login. Once you do, lower the opacity of the template, and place some pixels where there’s none

EDIT: Changed position to bottom right as it feels more natural to grow from the bottom up. I also increased its size/fidelity a bit since we had a bit more space there

0

Summer Solstice Style Competition

preview

If you like using the !aihorde@lemmy.dbzer0.com bot, do join the event and suggest your best options.

it would be nice to extend the competition to Lemmy somehow as most of our community is mostly active on discord. If anyone would be willing to help with that aspect, please let me know.

If you like using the !aihorde@lemmy.dbzer0.com bot, do join the event and suggest your best options.

it would be nice to extend the competition to Lemmy somehow as most of our community is mostly active on discord. If anyone would be willing to help with that aspect, please let me know.

0

Nivenly retroactively rejects the Haidra application

The Nivenly Foundation Haidra decision nivenly.org
preview

The Generative AI haters and luddites won the day and bullied Nivenly to remove Haidra, even without clear majority in opposition or a quorum.

I will write a post-mortem about this situation soonish.

The Generative AI haters and luddites won the day and bullied Nivenly to remove Haidra, even without clear majority in opposition or a quorum.

I will write a post-mortem about this situation soonish.

0

AI Horde Halloween style contest = free kudos!

Discord Join the A Division by Zer0 Discord Server! discord.gg
preview

It’s that time of the year again, Halloween is almost upon us, so it’s time for Haidra to come up with new styles! Head over the the event page and submit your best ideas.

The best suggestions will become official styles for the AI Horde and will be available in our bots! You also get kudos!


Contest rules:

Use any model on Horde. For a sd15 or sd21 model, it may include Loras and/or TIs, for SDXL it should not, unless the curated SDXL Lora list is already out and your Lora is on it.

  • First prize: 100k kudos (2 winners, one for SDXL and one for non-SDXL)
  • Second prize: 20k kudos (5 2nd prizes available)
  • Per-entry reward: 5k kudos
  • Entries per person: 3
  • Start: October 8 2023
  • End: October 30 2023, 5:00 PM PST

Click here to see examples of existing Halloween styles.

Remember that sampler and step count and cfg can all be changed if wanted (recommend keeping cfg lower or step count higher, in case the user adds other loras/TIs/bad prompts, but not required). Hires-fix is also available.

Each prompt should be accompanied by 2-5 images showing not just it’s strengths but also how widely it can be used (IE, don’t only show pictures of people if it also makes really good pictures of landscapes too)


Entries should be posted in Horde Discord at #Make a Halloween themed style if you have a Discord account. If you do not have a Discord account, you can post them here, but make sure to include your Horde ID (so I can send you kudos) and message me elsewhere (I don’t keep track of Lemmy well; you can message me on Reddit, a post on my steam profile matrix, or ping me in a GitHub comment).

Current entries are temporarily available in the AI-Horde-Styles repo, check the Halloween 2023 testing category; you can play with them for free in the Halloween 2023 style contest entries party on Discord.

It’s that time of the year again, Halloween is almost upon us, so it’s time for Haidra to come up with new styles! Head over the the event page and submit your best ideas.

The best suggestions will become official styles for the AI Horde and will be available in our bots! You also get kudos!


Contest rules:

Use any model on Horde. For a sd15 or sd21 model, it may include Loras and/or TIs, for SDXL it should not, unless the curated SDXL Lora list is already out and your Lora is on it.

  • First prize: 100k kudos (2 winners, one for SDXL and one for non-SDXL)
  • Second prize: 20k kudos (5 2nd prizes available)
  • Per-entry reward: 5k kudos
  • Entries per person: 3
  • Start: October 8 2023
  • End: October 30 2023, 5:00 PM PST

Click here to see examples of existing Halloween styles.

Remember that sampler and step count and cfg can all be changed if wanted (recommend keeping cfg lower or step count higher, in case the user adds other loras/TIs/bad prompts, but not required). Hires-fix is also available.

Each prompt should be accompanied by 2-5 images showing not just it’s strengths but also how widely it can be used (IE, don’t only show pictures of people if it also makes really good pictures of landscapes too)


Entries should be posted in Horde Discord at #Make a Halloween themed style if you have a Discord account. If you do not have a Discord account, you can post them here, but make sure to include your Horde ID (so I can send you kudos) and message me elsewhere (I don’t keep track of Lemmy well; you can message me on Reddit, a post on my steam profile matrix, or ping me in a GitHub comment).

Current entries are temporarily available in the AI-Horde-Styles repo, check the Halloween 2023 testing category; you can play with them for free in the Halloween 2023 style contest entries party on Discord.

0

Yesterday i discovered "AI Horde" [@aihorde](https://lemmy.dbzer0.com/c/aihorde), a crowdsourced distributed cluster of Image generation workers. You can try AI generation for free, no strings attache

preview

Yesterday i discovered “AI Horde” @aihorde, a crowdsourced distributed cluster of Image generation workers. You can try AI generation for free, no strings attached and even contribute yourself with your resources!

And I love the name, FOR THE HORDE! 😀

more info: https://aihorde.net/

#ai #generativeAI #aiart #aiimage #aihorde

Yesterday i discovered “AI Horde” @aihorde, a crowdsourced distributed cluster of Image generation workers. You can try AI generation for free, no strings attached and even contribute yourself with your resources!

And I love the name, FOR THE HORDE! 😀

more info: https://aihorde.net/

#ai #generativeAI #aiart #aiimage #aihorde

0

AI Horde birthday

preview

The AI Horde is about to become 1 year old, on Sep 3rd. I plan to make it a whole day event, where I’ll be on discord voice comms all day, just hanging out. You can ask me anything you want.

I hope people old and new join us to celebrate one year of crowdsourced Generative AI compute for everyone!

And everyone is welcome!

Discord Invite Link

The AI Horde is about to become 1 year old, on Sep 3rd. I plan to make it a whole day event, where I’ll be on discord voice comms all day, just hanging out. You can ask me anything you want.

I hope people old and new join us to celebrate one year of crowdsourced Generative AI compute for everyone!

And everyone is welcome!

Discord Invite Link

0

Open Discussions: Generative AI and Ethics of Data Collection.

Discord Discord - Group Chat That’s All Fun & Games discord.gg
preview

Al Horde First Open Discussions Event!

Generative AI and Ethics of Data Collection

Saturday, August 19 5pm PST

Hello everyone! As discussed previously, I’d like to host an open discussion related to the Al-Horde. Our first topic will be on the ethics of generative Al AND ethics of data collection. As for what to expect, a mostly unstructured only somewhat formatted discussion. Most likely an open stage channel where anyone can participate along with a text channel for those of us who are shy. Since this is the first of these events and level of participation among other factors is completely unknown, we might also move to a normal voice call with no agenda, script or direction at all. If this event does well, we will do more like it in the future.


Since R hasn’t posted anywhere other than Discord (that I know of), I figured I might as well post this here as well.

Al Horde First Open Discussions Event!

Generative AI and Ethics of Data Collection

Saturday, August 19 5pm PST

Hello everyone! As discussed previously, I’d like to host an open discussion related to the Al-Horde. Our first topic will be on the ethics of generative Al AND ethics of data collection. As for what to expect, a mostly unstructured only somewhat formatted discussion. Most likely an open stage channel where anyone can participate along with a text channel for those of us who are shy. Since this is the first of these events and level of participation among other factors is completely unknown, we might also move to a normal voice call with no agenda, script or direction at all. If this event does well, we will do more like it in the future.


Since R hasn’t posted anywhere other than Discord (that I know of), I figured I might as well post this here as well.

0

Feature request

preview

Hey,

I’m not sure if this kind of post fits this community or not, but I’ll give it a go. Is there a possibility we could add a bot which creates music? I’d be down to listen to some AI tunes during work, if we could find a fitting model. I have no idea if this is already a thing or not, I’m just sharing my thoughts.

Hey,

I’m not sure if this kind of post fits this community or not, but I’ll give it a go. Is there a possibility we could add a bot which creates music? I’d be down to listen to some AI tunes during work, if we could find a fitting model. I have no idea if this is already a thing or not, I’m just sharing my thoughts.

0

SDXL 1.0 Release Candidate now in rotation on the AI Horde

preview

We had the SDXL_beta available running for a few days. Previously you were getting images of SDXL 0.9 and SDXL 0.9 + Refiner. Now you’re going to get a comparison between SDXL 0.9 and SDXL 1.0 RC.

Please remember to tell us which image is the better one. Most frontends have been updated to allow this comparison. The above screenshot is a sample of how Artbot handles it.

Stability.ai is going to be regularly changing the models behind your generations so we’re going to be seeing better and better versions.

We had the SDXL_beta available running for a few days. Previously you were getting images of SDXL 0.9 and SDXL 0.9 + Refiner. Now you’re going to get a comparison between SDXL 0.9 and SDXL 1.0 RC.

Please remember to tell us which image is the better one. Most frontends have been updated to allow this comparison. The above screenshot is a sample of how Artbot handles it.

Stability.ai is going to be regularly changing the models behind your generations so we’re going to be seeing better and better versions.

0

Fun fact, you can follow our lemmy link aggregation by following this account [@aihorde](https://lemmy.dbzer0.com/c/aihorde) for AI Horde links and discussions or [@stable\_diffusion](https://lemmy.db

preview

Fun fact, you can follow our lemmy link aggregation by following this account @aihorde for AI Horde links and discussions or @stable_diffusion for #stablediffusion stuff. It’s like having a reddit feed in your mastodon follows. Only better because it’s #fediverse!

Fun fact, you can follow our lemmy link aggregation by following this account @aihorde for AI Horde links and discussions or @stable_diffusion for #stablediffusion stuff. It’s like having a reddit feed in your mastodon follows. Only better because it’s #fediverse!

0

[@aihorde](https://lemmy.dbzer0.com/c/aihorde) [@stablehorde\_generator](https://sigmoid.social/@stablehorde_generator) draw for me a horde of pirates raiding a computer mainframe style:piratepunk

preview

@aihorde @stablehorde_generator draw for me a horde of pirates raiding a computer mainframe style:piratepunk

@aihorde @stablehorde_generator draw for me a horde of pirates raiding a computer mainframe style:piratepunk

0

Competition event for making the new AI Horde and Haidra logos

Discord Join the A Division by Zer0 Discord Server! discord.gg
preview

Help us create something cool to using Stable Diffusion and the AI Horde. Selected entries get 100K kudos, public attribution and our gratitude!

Help us create something cool to using Stable Diffusion and the AI Horde. Selected entries get 100K kudos, public attribution and our gratitude!