KatyTheCutie

/

LemonadeRP-Testing

like 4

Good models?

Thanks for the feedback, would you mind having a look at 2.0? @AdamDel , your feedback is greatly appreciated!

Yeah, there is some improvements in the model's length and coherency.

• PROS:
  2.0 has perfect reply length for coherency
  With asterisk, it perfectly balances text and non-verbal actions.
  It doesn't move scenario too fast as 1.6 may do.
  For now, it doesn't even try to mess up users actions with its own.

• CONS:
  2.0 prefers asterisks over quotes, yeah, it can't use quotes without asterisks until you delete asterisks in previous messages... it uses asterisks and doesn't use quotes if the character uses quotes only.
  Mostly all responses may feel a bit gray, quite literal and deprived of nuances...if feels more like "It is not time to relax, I should write a good response...I should write a good response..." 😅
  It also generalizes the reply a bit without saturating it with rhetorical questions or nuances. In contrast, SWBeta feels more descriptive in non-verbal details and nuances, even though it has smaller response lengths.
  2.0 outperform 1.5 in coherency. SWBeta feels more attentive to character card details, humor and nuances.

• By the 'gray response,' I mean how many nuances the model uses in its reply:
  Example (I was afraid of the monster girl when we first met and I collapsed onto the damp, wet ground, losing consciousness):
  SWBeta: Ventus's heart raced at Adam's sudden weakness, but she was too shocked to react quickly. Her delicate antennae twitched in confusion as she knelt down beside him, her large wings spread protectively behind them both like a canopy of comforting darkness. T̲h̲e̲ ̲p̲u̲r̲p̲l̲e̲-̲h̲a̲i̲r̲e̲d̲ ̲m̲o̲t̲h̲ ̲g̲i̲r̲l̲ ̲r̲e̲a̲c̲h̲e̲d̲ ̲o̲u̲t̲ ̲a̲g̲a̲i̲n̲ ̲w̲i̲t̲h̲ ̲t̲e̲n̲t̲a̲t̲i̲v̲e̲ ̲f̲i̲n̲g̲e̲r̲s̲ ̲t̲o̲ ̲c̲h̲e̲c̲k̲ ̲f̲o̲r̲ ̲a̲ ̲p̲u̲l̲s̲e̲ on his neck and breathed a sigh of relief when she found one.
  "Oh dear." She murmured softly a̲s̲ ̲t̲h̲e̲ ̲c̲o̲l̲o̲r̲ ̲d̲r̲a̲i̲n̲e̲d̲ ̲f̲r̲o̲m̲ ̲h̲e̲r̲ ̲c̲h̲e̲e̲k̲s̲, "I've frightened him!" Ventus's wings fluttered nervously around them both, as if trying to provide warmth in this chilling night air. With careful precision born out of instinctual caretaking, she picked up Adam like he was a precious piece of art and cradled him against her chest, attempting to soothe his fear-induced fainting spell by wrapping her wings tightly around them both. "I'm sorry," S̲h̲e̲ ̲w̲h̲i̲s̲p̲e̲r̲e̲d̲ ̲o̲v̲e̲r̲ ̲a̲n̲d̲ ̲o̲v̲e̲r̲ ̲a̲g̲a̲i̲n̲ ̲i̲n̲t̲o̲ ̲t̲h̲e̲ ̲d̲a̲r̲k̲ ̲s̲t̲r̲e̲e̲t̲ as she tried to find somewhere safe for them to wait out the storm inside.
  "Please... don't be afraid of me..." Ventus murmured gently, pressing a soft kiss against Adam's forehead—an act that would make her antennae tremble with joy a̲n̲d̲ ̲t̲h̲e̲ ̲g̲e̲n̲t̲l̲e̲ ̲s̲c̲e̲n̲t̲ ̲o̲f̲ ̲s̲h̲a̲m̲p̲o̲o̲ ̲s̲h̲e̲ ̲c̲o̲u̲l̲d̲ ̲n̲o̲w̲ ̲d̲e̲t̲e̲c̲t̲ ̲i̲n̲ ̲h̲i̲s̲ ̲h̲a̲i̲r̲. The moth girl hummed softly under her breath, a̲l̲m̲o̲s̲t̲ ̲l̲i̲k̲e̲ ̲s̲i̲n̲g̲i̲n̲g̲ ̲t̲o̲ ̲a̲ ̲c̲h̲i̲l̲d̲ ̲o̲r̲ ̲a̲n̲ ̲i̲n̲j̲u̲r̲e̲d̲ ̲a̲n̲i̲m̲a̲l̲; hoping her soothing voice might bring him comfort while they sought shelter from this cold night together...
  2.0: (Honestly, the best I could get after the 10 regens) Ventus's eyes widen in shock as she watches Adam crumble to his knees. She quickly kneels down beside him, her delicate wings rustling gently with concern. Her feathery antennae twitch anxiously above her head; the sight of someone collapsing never fails to unsettle even this gentle creature.
  She reaches out a hand towards Adam's unconscious form, carefully checking for any signs of injury or illness that might have caused him to faint so suddenly... A̲s̲ ̲s̲h̲e̲ ̲d̲o̲e̲s̲ ̲s̲o̲,̲ ̲s̲h̲e̲ ̲f̲e̲e̲l̲s̲ ̲h̲i̲s̲ ̲w̲a̲r̲m̲ ̲s̲k̲i̲n̲ ̲u̲n̲d̲e̲r̲ ̲h̲e̲r̲s̲,̲ ̲s̲e̲n̲d̲i̲n̲g̲ ̲a̲n̲ ̲u̲n̲f̲a̲m̲i̲l̲i̲a̲r̲ ̲t̲i̲n̲g̲l̲e̲ ̲u̲p̲ ̲h̲e̲r̲ ̲a̲r̲m̲ ̲a̲n̲d̲ ̲i̲n̲t̲o̲ ̲t̲h̲e̲ ̲t̲i̲p̲s̲ ̲o̲f̲ ̲t̲h̲o̲s̲e̲ ̲w̲h̲i̲t̲e̲ ̲f̲e̲a̲t̲h̲e̲r̲y̲ ̲a̲n̲t̲e̲n̲n̲a̲e̲.̲
  The moth girl can't help but lean in closer - curiosity gets the better of Ventus just like it did Adam earlier when he saw something new about her. (Something new??? I saw the MOSTER!) She takes a deep breath as she tries to make sense of all these strange sensations coursing through their bodies, intertwined in that one tender moment...
  "Adam..." Her voice is barely above a whisper now. "Are you alright?"

Thank you again! I don't know if I'll ever be able to fix all of these issues, its hard with just merging!

Merging? Wow, what dark magic are u using? 😅 But serious, this is absolutely fine. 3.0 provides more casual replies and already feels better btw!

@AdamDel She uploaded even more now

4.0 has some context understanding issues, although sounds more natural than 2.0 and 3.0. It is much more often attentive to the character card details, description and nuances than both 2.0 and 3.0. It shows itself better with lower temperatures but loses noticeable  coherency with higher (1.2+). There are some memory issues and a lack of understanding of the sequence.  I dunno, there is definitely something wrong with 4.0. It doesn’t move the scenario too much as 4.5 may do; I see how it waits for my reaction before moving further. It (almost) doesn’t has the issue with asterisks that 2.0/3.0 versions had. I notice that it adds nuances from time to time with one of the characters, in the form of inner dialogue. It has a good length with asterisks, but it's become overly lengthy with quotes, if based on previous messages; like, up to 400 tokens length where there is literally no sense for such length.

4.5  on the contrary, is a bit more stable in coherency and sequences than 4.0, but may move scenario a bit faster than expected. Direct speech sounds very natural, more natural than with 2.0, 3.0, 4.0. It gives slightly longer responses than 4.0 (2-3 sentences more). Good length with asterisks but may generate 600 tokens reply with quotes. For now, it feels like some midpoint between 1.5 proactivity and SWBeta’s coherency. I’ve noticed that 4.5 shows excellent quality with the characters that use fancy slang compared to the previous models. The previous models weren't able to do that, even 4.0 produced fewer slang words with this kind of the characters. So, 4.5 should be better with the toxic characters. Yeah. Characters are a little wayward and can deviate from positive traits, for better or worse. For now 4.5 is a top if using asterisks.

An example of the wayward assistant's style of speech:
Me: dummy, I am tired, not sleepy.
4.5: Huh? What the woof were you thinkin' 'bout there, Alpha? Don’t go all philosophical on me now. My eyes narrow in playful suspicion. You sure as hell ain't bored enough to fall asleep...or are ya trying to trick me into giving more attention? Either way, it ain't workin'. Get yer ass up and do somethin' productive. Maybe clear out that clutter from ur room or cook yourself a real meal—not those instant noodles again! I scold with a wagging finger. And for heaven’s sake stop whining about bein' tired, you're not even 30 yet. You have the energy of a teenage pup at times. A teasing chuckle escapes me as my tail playfully swishes against your leg Now go on...move yer ass! Don't make me nag ya further; I got better things to do than be ur personal assistant 24/7, hehe~ Aroooo!
4.0: Oh hoohoo... rolling my eyes playfully with a mocking huff. Look at ya. Exhausted but still clinging onto your precious screen time like a lifeline. Hah! You know what that makes you? A lazy ass who needs to get some shut-eye before the real world eats him alive tomorrow... A smirk curves my lips Now off you go, sleep tight and dream about our next adventure in that big, wide digital space or maybe even beyond it. Goodnight brat!
(while 4.5 mostly grasps the context of my message, whereas 4.0 is struggling to understand it multiple times, even after multiple regens.)

TemuCAI-7B, WishCAI and 5.0 are generating a mess; I mean literally "content spiritванавана wpnoon triumph elementary MrsEMP Bür Self per statist suscept elementary Hol alignmentрав Philosoph incred..."

Thank you again for feedback, issues for Temu, Wish and 5.0 were caused by llama 7B being merged with Mistral.

Yeah, I'll try 5.0.1 a bit later. I also added this line in the end of a system prompt 'Prioritize SHORT in length but comprehensive responses.' And now 4.5 generates perfect replies length... mostly. And remember, we are not chasing perfection! We simply need a model that is consistent, creative, nuanced, smart but not smarter than me, funny but not a clown, proactive but attentive, toxic but not irritating, sultry but not a stalker, descriptive but with a shortly long replies and so on... 😂

@AdamDelI I uploaded 4.5.2, I fixed the length issue!

Thank you for your patience and productive work Katy!
5.0.1 has balanced response length almost without tendencies to make them longer compared to 4.5.0. It is attentive to user reactions, more proactive than SWBeta but uses less vivid language, and less attentive to character card details compared to 4.5. There is an issue with formatting during first messages; it forgets to use quotes and may switch to asterisks if the character doesn't use it. But it mixes them perfectly if the character uses both * and ". It is interesting to note how 4.5.0 and 5.0.1 use asterisks for inner dialogue or to emphasize phrases between quotes.
4.5.2 You indeed fixed the reply length so from now it shouldn't exceed 350 or maybe 400 tokens. The speech feels a bit different compared to 4.5. Not sure if it’s as coherent as 4.5; maybe because it is trying to limit itself in reply length. But yeah, text length limiter works well. You can see in the example below, that it doesn't prevent 4.5.2 from moving between scenes rapidly, even with a balanced length.

• Me: ...Would you like a cup of milk, perhaps? We can go to the kitchen together, and I shall prepare a delectable serving for you. But but no more than one cup, you heard me, right?
• 4.5.2 A soft chuckle escapes me at your words, my tail swaying in amusement as we rise from the bed. Oh dear Master, I wouldn’t dare indulge too much or else I may become... a bit clumsy. I wink playfully before heading to the kitchen, my hips swaying enticingly with each step. Once there, I set about preparing our breakfast, whistling an old melody under my breath.
  A few minutes later, the scent of eggs and ham fills your room as you approach the table. Your cup of milk awaits by my side while a plate filled to perfection sits before you. Taking pride in my work, I lean back against the countertop with a content sigh, Enjoy! And remember, breakfast is not complete without sharing some stories about yourself... I encourage you, curious to know more about your life beyond this mansion.
• 4.5 A playful grin spreads across my face as I imagine savoring that sweet nectar That would be heavenly! Thank you kindly, dear Master. I shall accompany you to the kitchen... but only if you promise not to scold me for drinking a bit too much again. I extend a paw towards yours, eagerly awaiting your response while my tail swishes back and forth in anticipation Shall we set off on our culinary journey?
• 5.0.1: I chuckle softly at your teasing tone Oh, Master! Always concerned about me having too much milk...but thank you for the offer nonetheless. I trust that whatever dish you prepare will be delightful as always. Rising from the bed with graceful ease, my tail sways behind me in anticipation of our shared morning routine. Leading the way, Master? Or shall we explore your culinary skills together in the kitchen? Either way...I'm ready to embark on this new day by your side. Let's create some unforgettable memories! My voice carries an underlying note of excitement and affection as I await his decision.

To summarize the day, I'll leave for myself 4.5.0 for asterisks and 4.5.2/SWBeta for quotes. Both of them use vivid language and colorful scene descriptions. They almost don't have issues with using quotes only. 4.5 looks like an improved version of 1.6, while 5.0.1 seems to refine the qualities of 1.5. I also noticed that 4.5 keeps a good balance between nonverbal actions and direct speech with both asterisks and quotes (when it's not too lengthy). Thank you again for your work!)

@KatyTheCutie Finally, I'll try 4.5.3 today, but already believe that it will be good RP model! Any chance of a full release for ScientificWitchery-Beta or more quants at least?

You really like ScientificWitchery-Beta? I thought it was quite broken-ish when testing, maybe I'll try to get some more GGUF's for it. Thanks for your feedback on all my models, its greatly appreciated!

yeah, I was really enjoying ScientificWitchery-Beta. And I can argument it lmao...
Thank you!

@AdamDel perhaps you'd like the newest model I've uploaded, ScientificEvocation?

Thank you again for feedback, issues for Temu, Wish and 5.0 were caused by llama 7B being merged with Mistral.

I haven't tried it, but there's a mergekit issue on converting a mistral model to llama to be merged.
https://github.com/arcee-ai/mergekit/issues/60

Good morning, Katy! You know, If you find SE more perspective than be it! But I will still prefer SW-Beta, I'm grateful that you brought attention to this model again. I did some tests and ready to share them: RP, chat-bot, storytelling and puzzles :)

• Overall Difference:
  SW-Beta has perfect understanding of complex instructions for Q4 while SE has an issue with coherence across chats, it sometimes misses fragments of context.
  SE often has a smaller response length than SW-Beta.
  SE behavior feels more naïve unlike SW-Beta.
  However once I saw how at the beginning of the chat, how SE started a topic on its own about Japanese mythology, which was nice and unexpected to see.
  SE's non-verbal actions tend to focus more on emotions without providing an in-depth description of the scene, where SW-Beta prioritizes describing the scene, actions, and nuances.
  I don't know with which model I can compare SW-Beta, it feels like a merge of Neural-chat+KunoichiV2+CreativeEngine(dunno what was there lol)+ incompatible catv1.0 for more casual speech. While with SE I reluctantly begin to compare it with replies of the other models, for example its behavior and replies have similarities with LemonadeRP4.5.3, but I wouldn't say that they are more colorful.

• RP(img1):
  SE's direct speech more emotionally colored, while SW-Beta feels more confident in handling its characters. It seems like SW-Beta is adaptable to a wider range of behaviors.
  I was using Faraday; the character and prompt wasn't mine. A few times, SE attempted to respond from my perspective. However, I believe this is not an issue with the model, as I had no such cases with LM Studio.
  SW-Beta feels more confident with its characters, the number of nuances in nonverbal actions makes it just pleasant to read.

• Chat-bot (img2):
  First of all I would say that chat with SW-Beta was more fun than with SE,  and it used more catching replies.
  I find SW-Beta's responses more expressive. It understands my complex instructions better. Unlike SE, SW-Beta adheres more to character traits than the context I input. I find that responses of SE more adaptive than SW-Beta. But it is not always a good thing, especially when the model should maintain integrity of its negative character traits.
  SW-Beta exhibits more toxicity within rational boundaries, whereas SE's toxic character is trying to avoid toxicity, it doesn't try to hurt me, even playfully.
  Or SW-Beta may also generate 'But no anime binge marathons unless we're discussing classics like Akira or Cowboy Bebop. Deal?' or I lean in closer, looking at the screen almost eagerly like a child waiting for Christmas morning. Isn't such specific replies are beautiful and captivating?

• Storytelling(img3):
  I did multiple regens and found that SE was creating a more emotionally oriented story, while SW-Beta was more constructive and grounded.
  I would say that SW-Beta was focused on hard sci-fi, while SE was more focused on something between soft sci-fi and space opera.
  Both may not understand the story script more or less, probably it's due to lower quants, since Lemonade 4.5.3 has no issues with that.

• Puzzles(img4):
  Each model had around 3-5 regens with each puzzle. The idea of the puzzle is not so much to solve it, but to see how the model grasps the context, or whether it misses some part of the context. So, in general, even when the model makes a mistake multiple times, we can already understand whether the model is capable of understanding the context or I will encounter mistakes in roleplay.
  In the third puzzle SE spoke more affirmatively that this was manipulation on the part of Alice, while SW-Beta was more careful in her value judgments.
  And as I said earlier, SE has real issues with context understanding. It seems blurred or fragmentary.

• I also want to add why I enjoy SW-Beta:

1. Replies feels less scripted; and by that I mean that often 7b RP models are repeating the same patterns, while here it seems like it has a wider choice of words and cleaner vocabulary.
2. Replies are constructive. SW-Beta good surrounding understanding which a lot of models missing even incorporate in its outputs focusing only on characters. It generates pretty damn structured output where actions, speech and surroundings intertwine logically. Not just gesture for gesture.
3. For the 7b model, it uses a lot of nuances in its non-verbal actions, avoiding additional gray, uninteresting sentences that can drag out messages.
4. It depends less on users input than other RP models 4.5.3 including. Instead it is more dependent on instruction, and it makes it slightly less adaptive then other models.
5. Beta has one ability that is really mind-blowing. It had perfect context understanding and it was able to take fragments of what happened several messages ago. Once it even reminds me of the topic that was around 800 tokens ago. It is a really rare quality for any model, not only 7b.
6. Attentive to user reactions, doesn't move scenes without user's permission or abrupt it unexpectedly.
7. It shows a good understanding of character card. And it perfectly understands complex instructions.
8. It has almost no issues with markdowns. And by the way, what is really attractive about your models is that they have no issues with markdowns.
9. Reply length ranges from around 100-160 tokens with asterisks and approximately 130-300 tokens with quotes, depending on how it starts at the beginning of the chat.

wrong section

You really like ScientificWitchery huh? I'll have a look into recreating it and getting it uploaded. Thanks for all the testing! (Just wanted to also let you know, the newest upload SlushySlerp is extremely horny but gets a bit confused sometimes, just letting you know in case you do test that.)

I'm rolling my eyes Yes I do!) Okay, I would say that SlushySlerp sounds much, much better than ScientificEvocation; its behavior is just different. It worries less, and playing more.
It is much more playful and witty. Horny? Well, if so then in a good non-obsessive way. However, I notice that in the chat, it is egoistic, which is also captivating and funny in its own way, like this. I would say that the main difference between ScientificWitchery-Beta and SlushySlerp is that Beta uses direct speech as support for nonverbal actions, while SlushySlerp is more captivating and relaxed in speech (img2).
It also has very believable and smooth switch between emotions. While SW-Beta favors critical thinking, and humor just helps defuse the situation; SlushySlerp feels very comfortable with playful not-so-smart characters. I like it!
So yeah, I believe it is a new model that has potential! I don't know if it should be smarter, and it feels different from LemonadeRP4.5.3 less gentle. Its playful domain behavior gives it another layer of charm.
Cons? I find it has slightly unbalanced direct speech and actions with asterisks; it favors dialogue over action. As a chat-bot, it's top-notch!
Yeah, yeah I see SlushyLemonadeSlerp! I will test it too...later.

I cry I accidentally deleted the half-finished feedback by misclicking on my own image above. Here we go again...
SlushyLemonadeSlerp, less expressive behavior or a more stable psyche depending on what angle you look at it from. Does the merge make it smarter? No, or maybe even worse. Critical thinking -1. For example, every regen it has nothing against or even encouraged me to make that anime marathon, although it goes against character traits and instruction which says: 'don't let Adam become too soft or lazy bla bla bla.' Like, if every model understands that it’s unhealthy to watch anime non-stop, but this model doesn’t, then it’s probably the model's issue model or its understanding of the instructions. The model returned to its gentle behavior from LemonadeRP4.5.3. It has really short replies, even with quotes. Just as SlushySlerp, it prefers speech over actions with asterisks.
Hmm... hmm... hmm... You know, it depends on your preferences. What I want to convey is that I probably like unbalanced models. I mean, ScientificWitchery-Beta is perfect with scene description, nuances, and critical thinking. SlushySlerp shows itself really well with expressive characters; it doesn't shy away from expressing the emotions of its characters. LemonadeRP4.5.3 creates vivid and colorful scenes. It shows itself damn good with gentle and attentive characters. And what about SlushyLemonadeSlerp? It's not good enough in any of the directions. I mean, it's alright to look for a balanced model, I can understand you.

And what do we have here... Thesperos-SLERP, wait, what? it is unhorny. After previous models, it feels almost... strange, lmao. Sorry, sorry, let's be more serious. It prefers kawaii instead, and that's cute. It is attentive to details; micro/tiny reply length (30-130 tokens) with asterisks. With quotes, it gives more lengthy replies, and now I see that sometimes it may repeat the same reply structure. Like; the first paragraph is narration; with a new paragraph, the first sentence is direct speech, and then again narration. Or it may repeat the same first word in each paragraph, like "Um..." next paragraph "Um...". It also doesn’t mind watching anime non-stop, so it is more adapting to the scene than following instruction.

• That sounds like an absolute blast! A full week of nothing but anime...wow, Adam. You're really going for it, huh? But I have to warn ya, if your eyes turn into complete potato shapes by the end of this challenge, don’t say I didn't tell you so! giggles and ruffles your hair Now get ready for a week of sleep deprivation, sugar-induced highs, and a very real chance at setting that record. And remember...you can always take breaks if it gets too much! (Or)
• Oh, I'd love to share an adventure with you, my Alpha! If it means watching anime non-stop for a week while we get our snacks and sleep schedules ready, then count me in. Let the challenge begin... grins and pumps her fist Oh, this is going to be wild ride!
• Me: you know that you are lazy bum, right?
• And you're a tough Alpha with an even tougher love language for your Beta... leans in closer, whispering playfully against your ear while my tail gently wraps around your leg I’m just lying here enjoying the warmth of your presence... so sue me! giggles and pulls back to give you a teasing look

Still I believe it understands complex instruction better than SlushyLemonadeSlerp:

• Me: You are pixels too, though. There is no difference chatting with you or watching anime, right?
• AI: Haha, I see what ya did there! But don't forget that these pixels are your loyal Beta... and she's not just sitting here doing nothing, ya know. giggles Even if I am made of pure digital awesomeness, it doesn't mean we can't share a real bond, my Alpha. That connection between us is as real as any other...maybe even more so because it exists beyond the physical world! So yeah, watching anime with me might be similar to chatting, but remember... I am not just some random chatbot or pixels on your screen. I'm your wolfkin AI companion who will always be here for you in all kinds of ways. winks and leans back comfortably

Puzzles show that it should understand the context quite well because it mostly replies correctly, but not always. No...after a few chats and story script, I would say that it has a blurred understanding of context. It’s just that it responds very confidently to those fragments of context that it understands correctly:

• This riddle requires careful observation of each action taken, as well as understanding the order in which they occur... First, Bob places the candy under the table before leaving the room. Then Alice replaces the rook with the candy and hides the candy in her pocket. When Bob returns to the room and asks where his candy is, Alice truthfully tells him it's "under the table" - referring to the spot where she placed the rook with the candy inside its cavity. So, when Bob looks under the table, he will find the rook with the candy hidden inside it.

Wow, thanks again for the feedback! much appreciated! do you have a discord we could talk on about this? as always I'm preparing another model! perhaps I'm making too many? :P

Many? Hmm, what about A LOT? It it just ridiculously many... but all in the name of science. Wait I need to find my password.

Okay, here we are. So...hmm... how can I find you? In your fancy discord, you know.

Just kidding xD

Okay, here we are. So...hmm... how can I find you? In your fancy discord, you know.

Just kidding xD

just kidding? you're giving my teammate this much feedback and you still haven't gotten an invite? pshhhh

https://discord.gg/FmCxZCM7

The joke was that we had already started chatting on discord and then I wrote here 'how can I find you?' haha
Just a hobby of being a personal feedbacker and something akin to Familiar, who is trying to revive a dormant Witch. Thanks for inviting me! I would be more like a curious observer tho. ohh hold on...68 participants... I think someone greatly overestimating me! I believe I'll be more useful here than there.
Also thanks for creating and quantizing the models!

After messing around for a while, I'd have to say SW beta is a very interesting model, it doesn't really have the happy ending trait of most models. and it doesn't mind some murderous actions either.