Welcome to Chatbot AI General #106, the thread for discussing and improving AI pony chatbots.▶ MLP Botshttps://mlpchag.neocities.orgSpreadsheet (CAI bots + Old repository):https://docs.google.com/spreadsheets/d/1J7BeqJVDS51cXF8Pgm2YZaFq-Z6ykSJTSillyTavern website extension: https://github.com/MLPChag/SillyTavern-Chag-SearchBunch of Character Cards (Descriptions/No Greetings):https://docs.google.com/spreadsheets/d/1Y6LNOCqAZAWIX-OBEv55HjlzcEdpUh-XbKCpbpL6v5kCAI bots converted to Tavern: https://files.catbox.moe/ckurq1.zipExpression packs: https://rentry.org/ChagExpressions!!!GALLERY!!!: https://drive.google.com/drive/u/2/folders/1Ao-h5HFGMPllSrzSBKM_BvGSiU9f0c2U▶ How do I start?1) Select a Frontend2) Select an AI model3) Select Jailbreak4) Select bots5) Lovemaking with AI mares!Starting in this hobby can be confusing and difficult. If it’s your first time and you’re lost,▶ ASK THE THREAD! ◀Novice-to-advanced guide: https://rentry.org/onrms▶ SillyTavern (preferred frontend)https://github.com/SillyTavern/SillyTavernOn Android: https://rentry.org/STAI-TermuxApp that voices pony responses in ST: https://drive.google.com/drive/folders/16Ss26VBmgzcSuTGzhaHqRuyVRceTf-YB▶ More frontends:Risu: https://risuai.xyzAgnai: https://agnai.chat▶ Localshttps://rentry.org/lunarmodelexperiments>>>/g/lmg/Mistral Nemo base model fine-tuned on fimfics: https://huggingface.co/Ada321/Nemo_Pony_2/tree/main▶ JailbreaksMLP JB: https://rentry.org/znon7vxeMore JB and guides: https://rentry.org/jb-listingHypebots for Tavern: https://rentry.org/pn3hb▶ BotmakingEditors: https://agnai.chat/editorGuides: https://rentry.org/meta_botmaking_listAdvanced: https://rentry.org/AdvancedCardWritingTricks▶ Miscellaneoushttps://rentry.org/ChagMiscellaneoushttps://rentry.org/ChagArchivePrevious thread: >>41775124**News:**- Deepseek releases V3 https://api-docs.deepseek.com/news/news1226- OpenAI announces o3 models, currently in early evaluation testing- Non-preview OpenAI o1 released on API- Gemini 2.0 Flash released, "more model sizes" in January https://developers.googleblog.com/en/the-next-chapter-of-the-gemini-era-for-developers/
Anchor for bots, lorebooks, scenarios.Previous >>41775127
Anchor for technical stuffs (Proxies, Updates, Models etc.)
Anchor for asking for bots, lorebooks, scenarios etc
Bots corner:MLP (With a full set of futuristic military gear and an AI assistant in your head, you find yourself in Equestria): https://mlpchag.neocities.org/view?card=MaudPie/MLP.pngKerfuffle: https://mlpchag.neocities.org/view?card=Anonymous/Kerfuffle.pngFriendship is Optimal UPDATE: https://mlpchag.neocities.org/view?card=Anonymous/Friendship%20is%20Optimal.pngArt corner:GPT-Pony: >>41803588Screenshots corner:Autism Council >>41775563Deepseek logs >>41777541, >>41776029, >>41777501 Candy cane >>41781103 Tummy war >>41781419, >>41781578 Horses can’t shower >>41784112
Screenshots corner 2:Do you sign? >>41782007, >>41782431 Teen Anon with AJ >>41786271 Cadance Found You >>41788098 Anon’s life >>41788334 Drunk Rara >>41789890 Hungry Anon >>41790996 Happy New Year >>41791870 WINGS? >>41792445
Screenshots corner 3:Pharynx >>41794041, >>41794049 Neighponese fan fiction >>41796003 Welcome to Wild >>41796567 Kerfuffle’s Backstory >>41798360 GPT-Pony VS Claude Pony >>41802604, >>41802898, >>41803466, >>41803725
Screenshots corner 4:Can they? >>41803760 Pinkie Bump >>41804738 GPT-Pony Smut >>41805446 Get Pregnant! >>41807099 Cute Kerfuffle >>41808496
>>41799613>>41808452
Lovemaking with AI mares!
>>41810417
>>>/g/103783087
I’m reading on /g/ that many Copilot accounts no longer work on Silly, showing: >message: Server Error Mine still works. Is anyone else having the same problem?
>>41810349Did they implement some kind of IP filter? I just made a new account. It let me connect to copilot and authorize, and the script even pretends to work, but I don't get access to any of the models."Debugging it" slightly by printing out the token in the get_token() shows that the script gets a 502 Bad Gateway. On ST's side, trying to connects prints "OpenAI status check failed. Either Access Token is incorrect or API endpoint is down.">>41810501I was literally writing this post as you posted.
>>41810505 Maybe try again or use another account? There are reports it was just temporary.>>>/g/103786638>>>/g/103786654>>>/g/103786679If it still doesn’t work, I can share my version of the script that works for me.
>>41810516Oh, you're right! It was broken for like half an hour, I think. Just some very unlucky timing.I am 99% sure it was on github's side since picrel is where I got 502, suggesting I wasn't reaching the endpoint in the first place. Maybe their server just died. idk, I'm not in webdev.
So pilot is actually fine? Can't check mine rn
>>41810537 Should be fine, yeah. Just a temporary problem on their side.
>>41810552<3
>>41809785 >How did you get it to react to filters/jailbreaks? The rejection in >>41803725 was a genuine rejection. I just decided to keep playing with it because it was fun in a meta sense. After that, I switched to a jailbreak I know usually allows me to do sexo.
Decided to check out this card again because of the discussion on advanced story cards from last thread
>>41810868God that rules.
>>41810614Neat, what model did you use?
>>41810928Chatgpt-4o-latest
>>41810317Google has updated code for Gemini Flash 2-0 and it now suddenly filters lots of folks. turns out it is nothing horrible just one thing needed to be adjusted in SillyTavern/proxyif you use Gemini Flash 2-0 then update SillyTavern to the latest Staging version >git switch staging>git pulland use Google AI Studio as source (not OpenAI). if you don't care about Flash 2-0 then ignore this post, it only applies to Flash 2-0, no other models were affectedmore info:https://github.com/SillyTavern/SillyTavern/issues/3268https://github.com/SillyTavern/SillyTavern/issues/3267and a commit if you want to patch your ST manuallyhttps://github.com/SillyTavern/SillyTavern/commit/2103e6238cbdf08232147edb165c26c32725dc02
Thoughts?
>>41811650Be nicer to Twilight.
>>41811650Can't say I get the appeal of incest fantasies but you're clearly having fun. I agree with the other guy though, be nice to your mare.
>>41810985Huh?
>>41811670How can that be incest?
>>41811686Adoptive incest fantasies are still incest fantasies. The parental relationship and the taboo that goes along with it is clearly an important part of that dynamic.
>>41811686>"If they knew their precious bookworm wanted to fuck the human colt she raised?"
>>41811682Nvidia had some giant releases, I accidentally posted on the wrong tab
>>41811698Isn't there a whole general here dedicated to that exact premise that says it isn't incest
>>41811712Yes and the average pornhub user looking up step-sister porn says it's not incest. They're incorrect.
Don't mind me
>>41811796What the fuck led to that?
>>41811798Coffeemaking with AI mares!
>>41811825Thank god there's a satisfying conclusion.
Is the github claude 3.5 really filtered somehow?I don't remember claude ever really refusing stuff before but now it seems extremely prone to blank responses and refusals.
>>41811828Let me fix that for you
>>41811856I assume you're the guy fucking around with teenage anon shenanigans?
>>41811849 I don’t find it more censored than on other sources. You just need to make sure your prefill is placed in a field like this at the end. There were some presets posted by anons if you want to try them: >"Otto preset modified for pony:"files.catbox.moe/w4bkp3.json >"Not a single refusal, even with darker stuff now:"files.catbox.moe/dz44qx.json >"CYOA RPG preset with the prefill in the correct place (read rentry.org/CharacterProvider-CYOARPG to see how it works)": files.catbox.moe/5c5zfc.json>>41811796Do you mind sharing your preset for those? I like them.
>>41811863Teenage anon? Anon is always 32 and balding in my stories
>>41811884I guess I incorrectly assumed that cumming in coffee is classic teenager behavior.
>>41811895Dafuq were you up to as a teen?
>>41811900Not saying I'd do it myself, I'm just saying that teenagers have very stupid, impulsive ideas that make zero sense to anyone other than themselves. Especially when they're upset.
>>41811904Spoken like someone who has jacked off into a coffee cup before
https://nvidianews.nvidia.com/news/nvidia-puts-grace-blackwell-on-every-desk-and-at-every-ai-developers-fingertips
>>41811965>128gb unified memory(VRAM equivalent?) for only 50% more than the 5090's 32gbwhy would anyone buy the GPUs for AI?
>>41811996 For pure AI, this does seem like a much better deal, yes. We just need to see the availability once it releases.
>>41811965 >Capable of running 200B-parameter models So for now, that would mostly mean running Mistral Large finetunes.
what do you need to run top of the line local? 405B?
>>41811996There's no free lunch. That thing has LPDDR5, it's many times slower than a GPU like 5090.If you're prototyping, these $3000 will give you a thousand hours on a very large server GPU on vast.ai. And you can spend it in a smart way - not everything has to be done on a big iron, something needs much bigger iron, many things can be pipelined to hell and back, something doesn't require a custom model and you can just pay a fraction of cost for tokens on a stock one, etc. With this $3000 thing you have neither flexibility to optimize your stuff, nor speed to do anything useful even in inference, let alone training.Basically it's Nvidia's answer to people buying 192GB macbooks to run heavyweight models, realizing they run at 5t/s and ending up writing a reddit post and putting it on the shelf. At least 5090 is fast enough to train image models and doubles as a gaming GPU.>VRAM equivalentlollmao
>>41811996>>41812272It will have good enough speeds for up to 405B models it looks like.https://www.theregister.com/2025/01/07/nvidia_project_digits_mini_pc/>From the renders shown to the press prior to the Monday night CES keynote at which Nvidia announced the box, the system appeared to feature six LPDDR5x modules. Assuming memory speeds of 8,800 MT/s we'd be looking at around 825GB/s of bandwidth which wouldn't be that far off from the 960GB/s of the RTX 6000 Ada. For a 200 billion parameter model, that'd work out to around eight tokens/sec.That would be about 4 tks for 405B, 8 for 200B, 20 for 70B
>>41812272And it looks like it will have about the bandwidth of a RTX 5070. It will be probably the best option if you want to run large models at affordable prices
>>41812307And I mention 405B because nvidia says they are built with high speed interconnect (they are basically cut down B200s) and you could use 2x of them for 405B. They are also capable and include tools for finetuning which needs a crap ton of compute / memory bandwidthProject DIGITS won’t run Windows: Instead, it will run DGX OS, a version of Ubuntu Linux customized with additional drivers and tools for developing and running AI applications. That’s the same software that runs on Nvidia DGX systems in the data center, meaning models built and tested locally on Project DIGITS can be deployed straight to the cloud or data center, the company said.Other Nvidia AI tools the device can run include orchestration tools, frameworks, and models on the Nvidia Developer portal and in its NGC. That includes the NeMo framework for fine-tuning models and the RAPIDS libraries for data science.
Also 5090 is said to be 104.8 TFLOPS (FP16)Digit is said to be 1 petaflop but at int4. So maybe 250 TFLOPS at FP16? I can't imagen they wont just not give it FP16 ability at all.
>>41811965Explain this in FiM terms
>>41810313https://files.catbox.moe/yli6w6.jsonlorebook based on >>41810137 and the posts within
>>41810313It's already in the dex, but it's been updated.https://www.characterhub.org/characters/Darkuzari/murky-waters-8c5be9d8b24a
Do you go with more than>{{user}} is the only human male in Equestriafor your persona?
>>41812809>not having multiple personas based on how seriously you're intending to take the chatfoolish
>>41812809>having a persona
>>41812809Yes because the persona prompt is very powerful and I can set up very funny results based on what I put in it. Next time you run a story put that you're a two pump chump in it
>>41812809Try this one. You won't regret it. Promise.
Haven’t tested any of them, but: >>>/g/103800108
>>41812891Kek.
>>41812891Now that's a proper character.
>>4181289110/10
>>41812809Yeah, I often tailor them to the card I'm playing, adding a few sentences to drive the scenario better. And of course establishing {{user}}'s age is really relevant, too. Ponies interact very differently to a 16 years old dumb punk and a graying 38 y.o. washed up boomer. Adding some appearance trait comes up, too.Think of it this way: Your persona is the second most important character for the model to know.
>>41426144Is this AI generated or just AI assisted?
>>41813439The image? I'm pretty sure it was 100% man-made. We received a lot of cool Chrysalis art for /chag/ by this anon.
>>41812809Just a mention of Human is all that's needed, not your special snowflake pandering.
>>41813480But I like being the alien ape the ponies are weirded out by.
>>41813484Based. The tension of ponies not being sure what to think of you is one of the best parts of any scenario involving a human in Equestria.
>>41813480>HumanYaght btw
>>41813480Have some respect for our resident artist newfag
>>41813911NTA but it's not related, anon. See the post quoted.
>>41813453>man-madeHeadcanon: it's literally Chrysalis shitposting here and making art, trying to harvest our love. Can barely type with her swiss cheese legs tho, hence the typos everywhere
>>41813736>You are going hell to
I don't know how you imitate fucking with hooves and forelegs. Seems tough.
>>41814617i feel so uncomfortable when i see fillies in anons logs
>>41814759Don't worry, this is simply a silly situation and they got told off.
>>41812665Updated.
>>41812906I can confirm that the first method works.
>>41815091 Nice. I still have SMS credits from when spamming the $18 free trial for GPT 3.5 Turbo was the meta.
>>41814759Okay, GPT.
>>41806092most of my logs with the card are more personal than i'd like to admit, but i do think it manages to mimic the weird way CelestAI talks pretty well. make sure you have example messages turned on, it helps
>>41815444and another,
>>41815444 I don’t know if it was by you too, but I liked this one.
chary deepseek status?
>>41815659haha glad to see at least one other autist use the card. that's perfect>>41810313friendship is optimal card update... againadded, fixed and tuned quite a few things. here's the link: https://files.catbox.moe/6uxhxk.pngwhile here i'd like to submit a lorebook. yes, i'm the anon who suggested it before, but i understand now that it'd be a pain to read the entire thread, put together a succinct description, name and image and upload it to the website. sorry for not seeing that earlier.>Lorebook name: Harmonizer>Image: Picrel>Description: Basic descriptions of notable characters and events in Equestria with certain lore tweaks to make the world feel more harmonious, so that every snowpity may shine.>Author: Anonapologies if it's a lot to put on your plate.https://files.catbox.moe/gv551t.json
>>41815814not coming, sorry. unsustainable
>>41812906Time to see if Opus is as good as I remember.
MM status?
>>41816045dead
>>41816045 MM funny joke. He’ll be back.
Looks like Chary lost the 11-20 4o GPT key for the time being. Time to work on cards again
>>41816142Fuck, hoping it comes back soon.At worst, public Unreliable has it.and >>41810469
>>41815825Why? I thought it was super cheap
>>41816045Just like >>41816082 said, it's not the first time he trolled like this. So I'm not worried (at least for now).
>>41816175 NTA, but it’s cheap for one person, not for 4,000.
>>41816142I'm so tired
>>41816142Did I miss something that we had 1120 exclusive cards?
>>41816201'Time to work on cards again' is someone just saying that they don't feel like they have any models worth genning with.
just saying but now would be the perfect time to deploy the boat
https://desuarchive.org/g/thread/103036011/#103036566
>>41816204Can't make a card to test if theres no models you can test it on.
>>41816216You can still take time writing greetings and starting the description, etc. But I agree that testing is necessary afterward to refine.At least there is still public GPT and Claude too if you can run the scripts.
mysterymare...
>>41816225I only make cards if there's models, otherwise I just take a break and go back to a hobby
>>41816142it is back-moderated endpoints are the same model but with extra Azure moderation - they work a bit faster than non-moderated models. if proxy is used for work (like coding) then consider to use -moderated for better "time to first token" but YMMV (-moderated models are deployed in more regions/servers, that's why)>>41816175not when 3700 people use it, yes
>>41816348Thanks
>>41816348>It's back
>>41816573That Equestria's a very chaste place if booping's considered pornographic.
>>41816573What model are you using?
>>41816601Look me in the eye and tell me someone booping your nose isn't a pretty intimate act.
>>41816676I'll look you in the eye and then boop you.
>>41816682That's it. You just booped your last snout.
>>41816699I regret nothing.
>>41816662Opus. Testing things since it's been a long while. I forgot how unhinged it could be with even just a slight push in the preset.
>>41816708Damn. I miss claude.
>>41816750There are several scripts posted, like the copilot to have 3.5.Opus one is more of a bother but worth it IMO.
>>41815824Updated, thanks anon.https://mlpchag.neocities.org/view?card=Anonymous/Friendship%20is%20Optimal.png
>>41815824And the Lorebook too. https://mlpchag.neocities.org/lorebooks>apologies if it's a lot to put on your plate. No problem. I was just rusty with adding lorebooks.
>>41815824Is Harmonizer meant more explicitly for FiO or is it just more of a generalized MLP lore pack?
>>41817142Generalized.
>>41817149Rad, thanks then, Anon.
The HuggingFace proxycalypse happened exactly one year ago.
>>41817169time indeed flies! wow
>>41817169What an ominous coincidence.
>>41817169what was the straw that broke the camels back?
>>41817203If I remember correctly, one spitefag went so hard on DDOSing HuggingFace proxies that the whole website crashed. They started removing proxies from then. I don't think it was related, but that was also soon after the million-dollar bill everyone raked up on the GPT-4-32K god/dev key.
>>41817209The spitefag problem was a sign /aicg/ was getting too big and attracted too much attention than a 4chan general should have.
>>41817232But /aicg/ always had spitefags. They are the reason why proxies were created on the first place, because they wouldn't stop revoking keys that were shared out in the open.
Does anyone know about whatever is happening in aicg right now? Should I be worried?
https://x.com/GeminiApp/status/1877039320147607581
>>41817378>Did you know that Gemini has a set of mini-experts called Gems>Gemsdarhling!
>>41817369>Does anyone know about whatever is happening in aicg right now? Fake proxy drama about a proxy (fiz one, don't think many are in it here). >Should I be worried? No. It's /aicg/ being /aicg/. If you've been here for a while, you know it's not the first time.
>>41817378 Unless I’m retarded and misunderstood, it’s like what they did with GPTs? Same models, but with some premade instructions baked in (e.g., check for spelling mistakes, etc.)? If that’s the case, I think the interest is limited since just prompting the base model gives better results.
>>41817423yes exactly this. they may finetuned those gems on custom tasks (maybe adapters like Apple did, or LoRAs) but I doubt. this service requires subscription - so they target your common John who is not familiar with prompting at all, it is not a product for the users who are already familiar with APIs
>>41816142>>41816348What's the endpoint of Charry again? The /openai one does not work.
>>41817435You don't need endpoints for Chary. You can stay with the default URL and just choose OpenAI as Chat completion. >>41810040 (Example is for Gemini, but you can keep the same URL for OAI)
>>41817402Alright. I’ve never been in that proxy, so I think I’m fine. Thanks.
>>41817453[proxy error (500): { "error": { "type": "proxy_internal_error", "proxy_note": "Reverse proxy encountered an error before it could reach the upstream API.", "message": "No GPT-4 keys available. Try selecting a Turbo model." }}]This is the notification. I'm using Agnai.
>>41817471 I get the same message in Tavern if I try to use Chorbo/Latte. Chary doesn’t have it. You need to switch to gpt-4o-2024-11-20 for a similar experience.
>>41817484Turns out I need to choose the "Openai with Chat and Template" one, not the usual Openai. Thanks for help.
Should MMbros be worried right now?
>>41817827Yes, you're all going to jail right meow.
>>41817827Probably not. If MM doesn't come back and he also doesn't say anything after a week, then there might be a reason to be worried.
>>41817827so what hapoened?
>>41810311starting messing around with deepseek but it seems like groupchats arent working for me. I started off with just talking with starlight and then added twilight. Now only twilight replies everytime while it just constantly shows starlight is typing when she tries to speak. Group chats of like 4 ponies was working for me earlier when running something local
I miss Mystery Mare...Chary is to slow with the queue times...
>>41818208Simply a reality of life. There will be times of abundance and times of famine.
>>41810313Up for rekindling an old flame? It's not the ai mare that you think.https://files.catbox.moe/5h88ah.png
>>41818127I was afraid seeing the flag, but that's a really cute log.
>>41818153I don't use Deepseek, so I can't help that much, but are you muting them and clicking on the dialogue bubble when you want one of them to talk? That's usually the preferred way to do group chat so it doesn't turn into a clusterfuck. Also, maybe try to update Tavern or make a second Tavern folder with an updated version to see if you still have the same problem. I know they put better support for Deepseek later on, notably with a way to directly select the model in Chat Completion
>>41818218Cool art.
>>41818245Thanks, I tried to proompt her pic at first but found slop wasn't cutting it wasting my time enough that I just did it myself.
>>41818258Based
>>41817203Everyone here got it wrong. The straw that broke the camel's back was the release of the Vulture article about us that caused chub.ai to go into full mass deletion mode.
>>41816209Literally how we're swimming in models
>>41812994How come I've never seen a black or Indian depiction of Anon before? He's always shown hwhite or smurf green.
>>41818341Because at the time Anon as a character was made users of this website were mostly white or smurf green themselves.
>>41818341>green The default and standard because the "Anonymous" pseudonym is green + greentext. The impossible color skin also works well to represent everyone since it's supposed to be a poster self-insert. >black or Indian 4chan + no one wants to represent him as such.
>>41813439Nah, no AI in tribute pictures. Each of them was made by hands fueled by pure gratitude and appreciation to that precious man who gives us access to expensive models free of charge. I know he deserves much more, but unfortunately that's all I can gib rn. (｡︶︵︶｡)>>41814136*autistic chirping*>>41818208Same... welp, at least today less things will distract me from finishing tribute pic that I wanted to make before new year holidays >.<
I'm glad for all those Claude scripts
>>41818241Usually I don't mute and just wait to see who responds first, but even with twilight muted it still has starlight stuck on showing that starlight is typing. hell even going back to previous group chats I've had to see if they work still now have the same problem where it has one character never respond, for example now it shows rainbow dash is typing but never responds
>>41818727Tried to search in the archive if someone had the same problem as you, but it doesn't seem so. I would try with a new Tavern folder up to date and do a quick test. If it works, it's then easy to just port all your cards and chats to the new one (at least on a computer).
I can't believe I am saying this but I already miss chorbo
>>41818833Pretty sure it will be back. Just take a pause or go with Copilot or other scripts for now if you can.
Coming back. Agnai user. Inside Column, Hopeman, and MM. What are my options?
>>41819060Hopeman has been dead for a while and won't come back. MM went down yesterday; I'm pretty sure it's temporary, but for now, it's out. Column is up and working right now. I'm just not sure if it's usable on Agnai, but if you could do it in the past, it should work. There are also scripts to run Claude without proxies (like Slaude in the past), but I'm not sure if they work with Agnai.
>>>/g/103827344>>>/g/103827226>>>/g/103827256
>>41819082What's available on Colly?
>>41819315>CollyColumn?Mostly Sonnet 3.5, Azure GPT too so gpt-11-20.
>>41818218Wanted to try it, but the card doesn't import in Silly. Anyone else?
>>41818218 >>41819603 There is no metadata, so unless it was just (a nice) art, anon fucked up.
>>41818833there's unreliable and cancer if you're okay with azure chorbo
>>41819082Is the claude on column scuffed?
>>41819656Do they have latest? Or just 11-20?
>>41819680 It uses another source, yes, but prefill and preset work. It goes up and down at random too. It worked not long ago, but now it seems to error. >>41819709Unreliable 11-20, Cancer has both I believe, but the password changed and you need to really read the proxy page.
>>41819643A shame. I really wanted some CAI mare.
>>41819722What's the link for cancer? I'd rather spend some time looking for a password than suffer with 11-20.
>>41819742 Pass and link changed, but same principle >>41816153 >suffer with 11-20.Uh, I quite like it.
how to delay github copilot ban? because i got banned about an hour after creating a new account
>>41819772Cant anymore. If you swipe more than like once or twice a min it will ban you.
>>41819772 >>41819799 Seems a bit random too from my experience. I went days on some and then just hours with the same rate of swipe.
>>41819799>>41819808do you mean I need to send less than two messages per minute? and I'm also wondering if it's possible to have NSFW chats without being banned? or only SFW?
>>41819827 I did heavy NSFW sessions and got banned only after five days recently, and got banned fast on SFW-only RP too, so I truly don't know what the ban conditions are. I switched to the script that asks for a phone number + smspool for now. It's a bit of a bother, but no ban, and I wanted to test Opus again.
>>41819827I was fine for over four days, got banned, then made a new account that got re-banned in like two hours when I was swiping more than usual.Then I made a third account, and that one has so far survived NSFW + high context, but I prompt/swipe pretty rarely.
>Tfw last I did this was in Silly 1.12 update>Feeling the need slowly rise up again after checking the logsLooking into this thread for old times was a mistake
>>41820093The ride never ends, lad. Chose a bad time to get the urge though.
>>41820098I can see, but I'll just be lurking around, seeing what I missed and reading up for now if there's no solution. And if there's one hopefully it still works, seeing how anons above are getting banned like the slaude era or something?No worries.
>>41820098>bad timeI think we're in the best spot in months. Even discounting closed proxies, there's unlimited free Sorbet, potential free Opus if you don't mind using a phone number, gemini (I quite like the model), open public chorbo proxies, pretty good locals, deepseek.Literally the only time things have been better were the free Opustopia days of endless public Opus, and Sorbet can be arguably just as good or better in some spots.>>41820121You can use email aliases to register on copilot (like "mail+1@mail.com") so bans are pretty inconsequential. And potentially avoidable, if the swiping thing is true.
>>41820121 >Slaude >>41810349 >>41812906 I won't say it's the best model, but you can also have Gemini. That's a fun model for free and without abuse >>41756306 There's also >>41816211 If you believe you will stay a while, it may also be worth it to read the Novice guide until the end.
>Do pony RP >AI decides on a random swipe to tell the druggy adventure of "Johnny Chainsaw" and "Fast Eddie" Alright... all the other swipes are fine and normal too.
>>41820181What, you're not interested?
>>41820207 Kind of... Spoilering it since it's not a pony log despite my attempts.
Does anyone here make their own models and implement papers?
>>41820276 Nope. Best we got was an Anon finetuning Mistral Nemo on the Fimfic Archive. https://huggingface.co/Ada321/Nemo_Pony_2/tree/main
Is column working for anyone else right now or are errors coming up? Internal server error?
>>41820135Oh there's public chorbo proxies? You know if you can point me at them, my last proxy poofed a bit back.
>>41820329 Yep, it’s fucked up right now. >>41820332 Just read a few posts above.
>>41820341Ahh... Yeah, it's right there. Sorry, still a little sleepy so I'm a bit stupid. Thanks for the help though.
>>41820300Sad. Fimfarchive has few garbage stories as it was mentioned in >>41498865, I hope they were removed from training dataset.
>>41820424 I’d suggest maybe checking /lmg/, but I don’t think anyone has made their own models there either. It’s mostly finetunes and merges.
>>41820341Alright. I just wanted to make sure my token still works.
>>41820135>free OpusWhere?
>>41818341I want someone to generate a story about Indian anon trying to woo Celestia in broken English
>>41820527It's linked at the top of the thread, but the $10 will probably run out quick with Opus.
>>41818757thinking its just a problem with deepseek, made a new tavern folder and it worked for like 2 messages in the group chat but ended up with the same problem, tried switching over to something like gpt 4o-mini and group chat worked right away
>>41819772We don't know how tf it works yet. But, the last time, after I was banned instantly for like 2 times in a row (noswipes, about 10k of context, sfw chat), fren assumed that it may have something to do with your browser/ip footprint. Next acc I've made using another browser+stable vpn(both for registration and talking to my little harem in tavern) didn't get banned yet after 5 days of my schizoRPs. Let’s see how long it will last.By the way, there's rumors about paid-pilot version being immune to ban waves. I don't know if it's true, just telling frens what I've heard around. (If someone here is rich enough to try, please share your experience.) @_@ +just a reminder to always disable all the "sharing data" options in pilot's settings. frens are forgetting sometimes. sorry for esl mistakes. im too eepy to recheck this text.
>>41820276https://huggingface.co/setfunctionenvironment/silverspoon-v1-72b
>>41821126Oh? Guess I'll need to try it.
>>41821139it kinda sucks I'm making a v2
>>41819643>>41819603>>41819724Funny, she definitely exists but somehow got corrupted in catbox when I try the link. The jews did this.>>418103132nd pass on her, so trying again. for collabs this would be a good one to try to improvehttps://files.catbox.moe/w8ufmy.png
>>41821147Did it actually improve its pony knowledge?
>>41821160well the dataset had1. Multiturn dataset directly from mlp Fandom2. Mlp general data from internet3. Mlp character card roleplay dataset4. autocomplete from all fimfiction books with like 20+ likesunfortunately something went wrong with the training I think
>>41821156It works for me now and looks like a neat card. Thanks.
