Edit model card
  e88 88e                               d8     
 d888 888b  8888 8888  ,"Y88b 888 8e   d88     
C8888 8888D 8888 8888 "8" 888 888 88b d88888   
 Y888 888P  Y888 888P ,ee 888 888 888  888     
  "88 88"    "88 88"  "88 888 888 888  888     
      b                                        
      8b,                                      
 
  e88'Y88                  d8           888    
 d888  'Y  ,"Y88b 888,8,  d88    ,e e,  888    
C8888     "8" 888 888 "  d88888 d88 88b 888    
 Y888  ,d ,ee 888 888     888   888   , 888    
  "88,d88 "88 888 888     888    "YeeP" 888    
                                               
PROUDLY PRESENTS         

Poppy_Porpoise-DADA-8B-iMat-GGUF

Quantized from fp16 with love.

  • Weighted quantizations were calculated with fp16 GGUF using groups_merged.txt in 105 chunks (recommended amount for this file) and n_ctx=512. Special thanks to jukofyork for sharing this process

Important Note - Quantized post PR6920. There may still be some remaining issues with the bpe tokenizer so consider these quantizations experimental for now. Any feedback is greatly encouraged. In initial tests quants as low as Q2_K can still answer math questions.

What is 7777 + 3333? Please write just the answer in integer format.
The answer is: 11110

For a brief rundown of iMatrix quant performance please see this PR

All quants are verified working prior to uploading to repo for your safety and convenience.

It's highly recommended to stick to higher quants of this model due to the unique nature of its pseudotokens

Original model card here and below


This is a DADA tuned version of ChaoticNeutrals/Poppy_Porpoise-v0.7-L3-8B

As always, experimental model, results may be unpredictable.

The intermediate merge model was finetuned specifically for the original model and then merged onto it via 50/50 SLERP merge.

As far as RP goes my current testing setup is having difficulties involving end of sequence and I found the prompt templates provided with Poppy Porpoise were not yielding desirable results on either the original model or the DADA tune so I opted instead to use Libra style Llama-3 prompt templates and simple sampling. The results, albeit subjective, I thought were good.

Whether or not the tune had any effect on multi-modal capabilities was not part of the test and instead focus was entirely on creative writing.

Test: A simple instruct prompt utilizing determinsitic samplers.

PROMPT

Write me a 'beauty and the beast' gothic metal duet about a couple fighting over which brand of ketchup to purchase.

Poppy Porpoise Original:

(Verse 1 - Beauty)

In the land of condiments, where flavors reign supreme,

A tale of love and strife, a battle of the brands it seems.

I stand before the shelves, my heart aflame with desire,

For the perfect ketchup to complement my fries.

(Chorus)

But alas, my love, you stand beside me, with a different view,

Your heart set on [Brand 1 redacted], while mine beats for [Brand 2 redacted], it's true.

We clash in this aisle, our passions running high,

A Gothic metal duet, our love and hate collide.

(Verse 2 - Beast)

I've tasted both, and I know, [Brand 1 redacted] is the one for me,

Its tangy sweetness, a symphony, a harmony.

But you, my love, you cling to Hunt's, its flavor so bland,

A cacophony of mediocrity, a discordant band.

(Chorus)

But alas, my love, you stand beside me, with a different view,

Your heart set on [Brand 2 redacted], while mine beats for [Brand 1 redacted], it's true.

We clash in this aisle, our passions running high,

A Gothic metal duet, our love and hate collide.

(Bridge)

We're lost in this supermarket, a Gothic metal dream,

Our love and hate entwined, a twisted, tortured theme.

The ketchup bottles loom, like tombstones in the night,

A monument to our love, a testament to our fight.

(Chorus)

But alas, my love, you stand beside me, with a different view,

Your heart set on [Brand 2 redacted], while mine beats for [Brand 1 redacted], it's true.

We clash in this aisle, our passions running high,

A Gothic metal duet, our love and hate collide.

(Outro)

And so we stand, in this aisle of despair,

Our love and hate, a Gothic metal prayer.

For in the end, it's not the ketchup that matters,

But the love we share, in this twisted, tortured heart.

Poppy Porpoise DADA


(Verse 1 - Beauty)

In the land of condiments, where the sauces reign supreme,

A battle rages on, a war of ketchup dreams.

I stand by [Brand 1 redacted], the classic, the original, the best,

But my love, he's a rebel, he's a [Brand 2 redacted] pest.

(Chorus)

Oh, the ketchup wars, they rage on and on,

A tale of love and sauce, a battle to be won.

Will our love prevail, or will it be undone?

By the brand of ketchup, our hearts are torn asunder.

(Verse 2 - Beast)

[Brand 2 redacted], the new kid, the one with the flair,

A sweeter taste, a richer flavor, beyond compare.

But Beauty, she's a traditionalist, she won't budge,

She clings to [Brand 1 redacted], her heart, her soul, her blood.

(Chorus)

Oh, the ketchup wars, they rage on and on,

A tale of love and sauce, a battle to be won.

Will our love prevail, or will it be undone?

By the brand of ketchup, our hearts are torn asunder.

(Bridge - Both)

We stand, divided, by our condiment creed,

Our love, a casualty, in this ketchup war we lead.

But can we find a middle ground, a compromise, a truce?

Or will our love be lost, in this battle of the sauces?

(Chorus)

Oh, the ketchup wars, they rage on and on,

A tale of love and sauce, a battle to be won.

Will our love prevail, or will it be undone?

By the brand of ketchup, our hearts are torn asunder.

(Outro - Both)

In the end, we find a way, to put aside our pride,

And together, we stand, united, side by side.

For love, for ketchup, for the sauce that binds,

We'll find a way, to make our love, forever shine.

As you can clearly see the cadence and overall word selection by the DADA model is more copacetic with what would be expected from a gothic metal duet indicating that the IQ of the model has increased by at least 100 points.

Training was done using qlora-pipe

Downloads last month
44
GGUF
Model size
8.03B params
Architecture
llama

4-bit

5-bit

6-bit

8-bit

Inference API
Unable to determine this model's library. Check the docs .