3 ms·

Tangential: I've set up a Discord Bot that turns your text prompt into images using Stable Diffusion. You can invite the bot to your server via https://discord

Tangential: I've set up a Discord Bot that turns your text prompt into images using Stable Diffusion. You can invite the bot to your server via https://discord.com/api/oauth2/authorize?client_id=1013373043625705513&permissions=117760&scope=bot%20applications.commands https://discord.com/api/oauth2/authorize?client_id=101337304... Talk to it using the /draw Slash Command. It's very much a quick weekend hack, so no guarantees whatsoever. Not sure how long I can afford the AWS g4dn instance, so get it while it's hot. Oh and get your prompt ideas from https://lexica.art https://lexica.art if you want good results. PS: Anyone knows where to host reliable NVIDIA-equipped VMs at a reasonable price?

olladecarne 4y ago
One thing I noticed is that on GCP if you create a a2-ultragpu (Nvidia a100 80gb) and you select a spot instance, the price estimate goes down to $0.33 hourly ($240/m) which sounds really good if it's not a mistake. I was wondering if you could then turn a single A100 into 7 GPUs using Multi-instance GPUs. So on an 80gb one you get 7 10GB GPUs (can't have 8 due to yield issues on those cards). I'm pretty sure that will run much slower than on the full instance, but not 7x slower so if you're running a larger service at scale this could be an option to parallelize things. If someone is able to get that running please let me know how it performs. The next thing I considered was just buying up a ton of 3060 12gb cards (saw a few new ones for $330) and just hosting a server from my house. This might be a good option if you don't care about speed but care about throughput. RTX 3090s are also decent in terms of price per iteration of Stable Diffusion. If you want to build a fast service like Dreamstudio I think it's the only option to be able to do it at a reasonable price. If you want to host these in the cloud using consumer RTX cards, you'll have to go with less reputable hosts since Nvidia doesn't allow it. I don't want to name any since I can't vouch for them, but there are some if you search. The cheapest option will be to buy them and host it yourself. I'm still researching what the best price/performance is for hosting this so if you have any findings please share.
rexreed 4y ago
I'm experimenting with your Discord bot right now. It would be great to have a command that shows where your processes are currently in the queue or maybe the discord bot can update on queue position.
- ManuelKiessling 4y ago
  Good idea, I'll look into it.
  - rexreed 4y ago
    I submitted 2 /draw requests with prompts, got quoted a time 15-30 min for first one and then 17-34 for 2nd, submitted about 5 minutes apart but it's been now past the upper limit of the quoted time without any results. I'm assuming that the image generation has failed or perhaps the bot got stuck. Having some way of knowing would be helpful.
    - ManuelKiessling 4y ago
      Not stuck, just a full queue. Results will come back sooner or later. Time is really just a guesstimate.
      - rexreed 4y ago
        Just got one of the images back. Looks like you might want to double your time estimates. Also I got a Rick Roll meme image back as one of the results. I assume this is some sort of failure mode response?
        ManuelKiessling 4y ago
        Rick chimes in when the AI thinks the image might be NSFW.
        rexreed 4y ago
        Well that would have been a fun interpretation of my prompt ;)
        hidelooktropic 4y ago
        I got this with the prompt "A cartoon backdrop of an empty hospital bed surrounded by medical equipment" Racy.
nabakin 4y ago
Any chance of releasing the source? I'd like to host my own instance so my discord server doesn't have to worry about queue times
- ManuelKiessling 4y ago
  Yeah, sure: https://github.com/manuelkiessling/stable-diffusion-discord-bot https://github.com/manuelkiessling/stable-diffusion-discord-... I quickly polished things and created a useful README - hopefully it's all correct. If not, let me know!
  - nabakin 4y ago
    Awesome, tysm!
    - ManuelKiessling 4y ago
      Pretty sure this one is important: https://github.com/manuelkiessling/stable-diffusion-discord-bot/issues/1 https://github.com/manuelkiessling/stable-diffusion-discord-...
    - ManuelKiessling 4y ago
      Consider joining our Discord server if you would like to discuss the project further: https://discord.gg/nsfeutx35z https://discord.gg/nsfeutx35z

olladecarne 4y ago

One thing I noticed is that on GCP if you create a a2-ultragpu (Nvidia a100 80gb) and you select a spot instance, the price estimate goes down to $0.33 hourly ($240/m) which sounds really good if it's not a mistake. I was wondering if you could then turn a single A100 into 7 GPUs using Multi-instance GPUs. So on an 80gb one you get 7 10GB GPUs (can't have 8 due to yield issues on those cards). I'm pretty sure that will run much slower than on the full instance, but not 7x slower so if you're running a larger service at scale this could be an option to parallelize things. If someone is able to get that running please let me know how it performs. The next thing I considered was just buying up a ton of 3060 12gb cards (saw a few new ones for $330) and just hosting a server from my house. This might be a good option if you don't care about speed but care about throughput. RTX 3090s are also decent in terms of price per iteration of Stable Diffusion. If you want to build a fast service like Dreamstudio I think it's the only option to be able to do it at a reasonable price. If you want to host these in the cloud using consumer RTX cards, you'll have to go with less reputable hosts since Nvidia doesn't allow it. I don't want to name any since I can't vouch for them, but there are some if you search. The cheapest option will be to buy them and host it yourself. I'm still researching what the best price/performance is for hosting this so if you have any findings please share.

rexreed 4y ago

I'm experimenting with your Discord bot right now. It would be great to have a command that shows where your processes are currently in the queue or maybe the discord bot can update on queue position.

ManuelKiessling 4y ago
Good idea, I'll look into it.
- rexreed 4y ago
  I submitted 2 /draw requests with prompts, got quoted a time 15-30 min for first one and then 17-34 for 2nd, submitted about 5 minutes apart but it's been now past the upper limit of the quoted time without any results. I'm assuming that the image generation has failed or perhaps the bot got stuck. Having some way of knowing would be helpful.
  - ManuelKiessling 4y ago
    Not stuck, just a full queue. Results will come back sooner or later. Time is really just a guesstimate.
    - rexreed 4y ago
      Just got one of the images back. Looks like you might want to double your time estimates. Also I got a Rick Roll meme image back as one of the results. I assume this is some sort of failure mode response?
      - ManuelKiessling 4y ago
        Rick chimes in when the AI thinks the image might be NSFW.
        rexreed 4y ago
        Well that would have been a fun interpretation of my prompt ;)
        hidelooktropic 4y ago
        I got this with the prompt "A cartoon backdrop of an empty hospital bed surrounded by medical equipment" Racy.

nabakin 4y ago

Any chance of releasing the source? I'd like to host my own instance so my discord server doesn't have to worry about queue times

ManuelKiessling 4y ago
Yeah, sure: https://github.com/manuelkiessling/stable-diffusion-discord-bot https://github.com/manuelkiessling/stable-diffusion-discord-... I quickly polished things and created a useful README - hopefully it's all correct. If not, let me know!
- nabakin 4y ago
  Awesome, tysm!
  - ManuelKiessling 4y ago
    Pretty sure this one is important: https://github.com/manuelkiessling/stable-diffusion-discord-bot/issues/1 https://github.com/manuelkiessling/stable-diffusion-discord-...
  - ManuelKiessling 4y ago
    Consider joining our Discord server if you would like to discuss the project further: https://discord.gg/nsfeutx35z https://discord.gg/nsfeutx35z