QuantFactory/L3-Super-Nova-RP-8B-GGUF

This is quantized version of Casual-Autopsy/L3-Super-Nova-RP-8B created using llama.cpp

Original Model Card

Image generated by mayonays_on_toast - Sauce



L3-Super-Nova-RP-8B

This is a role-playing model designed with the goal of good creativity and intelligence to improve advance role-playing experiences. The aim of L3-Super-Nova-RP-8B is to be good at Chain-of-Thoughts, summarizing information, and recognizing emotions. It also includes data about the human body and mind in an attempt to enhance understanding and interaction within role-playing scenarios.

The model was developed using various methods in multiple merging steps. To boost creativity, it used techniques to strengthen and adjust its output which was paried with the newly released merge method. All merge calculations were done in float32 format and then converted to the usual bfloat16 during merging.



Presets


Text Gen

The current good starting preset for this model. Subject to change. Settings by yours truly

Top K: 40
Min P: 0.075 # I've got some good results as low as 0.05 as well
Repetition Penalty: 1.01
# Don't make this higher, DRY handles the bulk of Squashing Repetition.
# This is just to lightly nudge the bot to move the plot forward
Rep Pen Range: 2048 # Don't make this higher either.
Presence Penalty: 0.03 # Minor encouragement to use synonyms. Don't make this higher maybe?
Smoothing Factor: 0.3

DRY Repetition Penalty:
  Multiplier: 0.8
  Base: 1.75
  Allowed Length: 2
  Penalty Range: 4096

Dynamic Temperature:
  Min Temp: 0.5
  Max Temp: 1.25
  Exponent: 0.85

Context/Instruct

Virt-io's SillyTavern Presets work really well with this.



Usage Info

Some of the INT models were chosen with some of SillyTavern's features in mind, such as emotion based sprites, dynamic music, and pretty much any feature, extension, or STscript that uses sumarization. With that said, it's recommended to use SillyTavern as your front-end.

While not required, I'd recommend building the story string prompt with Lorebooks rather than using the Advance Formatting menu. The only thing you really need in the Story String prompt within Advance Formatting is the system prompt. Doing it this way tends to keep the character more consistent as the RP goes on as all character card info is locked to a certain depth rather than getting further and further away within the context.



Quants

GGUF:

Exl2:



Merge Info

The merge methods used were Ties, Dare Ties, Breadcrumbs Ties, SLERP, and DELLA.

The model was finished off with both Merge Densification, and Negative Weighting techniques to boost creativity.

All merging steps had the merge calculations done in float32 and were output as bfloat16.


Models Merged

The following models were used to make this merge:



Evaluation Results


Open LLM Leaderboard

Explaination for AI RP newbies: IFEval is the most important evaluation for RP AIs as it determines how well it can follow OOC, Lorebooks, and most importantly character cards. The rest don't matter. At least not nearly as much as IFEval.

Metric Value
Avg. N/A
IFEval (0-Shot) N/A
BBH (3-Shot) N/A
MATH Lvl 5 (4-Shot) N/A
GPQA (0-shot) N/A
MuSR (0-shot) N/A
MMLU-PRO (5-shot) N/A

UGI Leaderboard

Information about the metrics can be found at the bottom of the UGI Leaderboard in the respective tabs.

Metric(UGI-Leaderboard) Value Value Metric(Writing Style)
UGI(Avg.) 23.56 0.199 RegV1
W/10 5.8 0.218 RegV2
Unruly 22.5 0.15 MyScore
Internet 11.8 8.34 ASSS
Stats 18.7 10.26 SMOG
Writing 31.5 1.76 Yule
PolContro 33.3


Secret Sauce

The following YAML configs were used to make this merge.


Super-Nova-CRE_pt.1

models:
  - model: nothingiisreal/L3-8B-Celeste-v1
  - model: Nitral-AI/Hathor_Tahsin-L3-8B-v0.85
    parameters:
      density: [0.35, 0.45, 0.5, 0.55, 0.65, 0.55, 0.5, 0.45, 0.35]
      weight: [0.495, 0.165, 0.165, 0.495, 0.495, 0.165, 0.165, 0.495]
  - model: Sao10K/L3-8B-Stheno-v3.2
    parameters:
      density: [0.65, 0.55, 0.5, 0.45, 0.35, 0.45, 0.5, 0.55, 0.65]
      weight: [0.165, 0.495, 0.495, 0.165, 0.165, 0.495, 0.495, 0.165]
merge_method: dare_ties
base_model: nothingiisreal/L3-8B-Celeste-v1
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-CRE_pt.2

models:
  - model: nothingiisreal/L3-8B-Celeste-v1
  - model: ChaoticNeutrals/Poppy_Porpoise-1.0-L3-8B
    parameters:
      density: [0.35, 0.45, 0.5, 0.55, 0.65, 0.55, 0.5, 0.45, 0.35]
      weight: [0.165, 0.495, 0.495, 0.165, 0.165, 0.495, 0.495, 0.165]
  - model: Sao10K/L3-8B-Lunaris-v1
    parameters:
      density: [0.65, 0.55, 0.5, 0.45, 0.35, 0.45, 0.5, 0.55, 0.65]
      weight: [0.495, 0.165, 0.165, 0.495, 0.495, 0.165, 0.165, 0.495]
merge_method: dare_ties
base_model: nothingiisreal/L3-8B-Celeste-v1
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-UNC_pt.1

models:
  - model: turboderp/llama3-turbcat-instruct-8b
  - model: ChaoticNeutrals/Domain-Fusion-L3-8B
    parameters:
      density: 0.5
      weight: [0.495, 0.165, 0.165, 0.495, 0.495, 0.165, 0.165, 0.495]
  - model: migtissera/Llama-3-8B-Synthia-v3.5
    parameters:
      density: 0.5
      weight: [0.165, 0.495, 0.495, 0.165, 0.165, 0.495, 0.495, 0.165]
merge_method: dare_ties
base_model: turboderp/llama3-turbcat-instruct-8b
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-UNC_pt.2

models:
  - model: turboderp/llama3-turbcat-instruct-8b
  - model: TheDrummer/Llama-3SOME-8B-v2
    parameters:
      density: 0.5
      weight: [0.165, 0.495, 0.495, 0.165, 0.165, 0.495, 0.495, 0.165]
  - model: ChaoticNeutrals/Hathor_RP-v.01-L3-8B
    parameters:
      density: 0.5
      weight: [0.495, 0.165, 0.165, 0.495, 0.495, 0.165, 0.165, 0.495]
merge_method: dare_ties
base_model: turboderp/llama3-turbcat-instruct-8b
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-INT_pt.1

models:
  - model: TheSkullery/llama-3-cat-8b-instruct-v1
  - model: FPHam/L3-8B-Everything-COT
    parameters: 
      density: 0.5
      weight: [0.139, 0.139, 0.208, 0.139, 0.208]
  - model: Ayush-1722/Meta-Llama-3-8B-Instruct-Summarize-v0.2-24K-LoRANET-Merged
    parameters: 
      density: 0.5
      weight: [0.139, 0.208, 0.139, 0.208, 0.139]
  - model: OEvortex/Emotional-llama-8B
    parameters: 
      density: 0.5
      weight: [0.208, 0.139, 0.208, 0.139, 0.139]
  - model: lighteternal/Llama3-merge-biomed-8b
    parameters: 
      density: 0.5
      weight: [0.208, 0.139, 0.139, 0.139, 0.208]
  - model: Casual-Autopsy/Llama3-merge-psychotherapy-8b
    parameters: 
      density: 0.5
      weight: [0.139, 0.208, 0.139, 0.208, 0.139]
merge_method: ties
base_model: TheSkullery/llama-3-cat-8b-instruct-v1
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-INT_pt.2

models:
  - model: TheSkullery/llama-3-cat-8b-instruct-v1
  - model: FPHam/L3-8B-Everything-COT
    parameters:
      density: 0.9
        gamma: 0.01
        weight: [0.139, 0.208, 0.208, 0.139, 0.139]
  - model: Ayush-1722/Meta-Llama-3-8B-Instruct-Summarize-v0.2-24K-LoRANET-Merged
    parameters:
      density: 0.9
      gamma: 0.01
      weight: [0.208, 0.139, 0.139, 0.139, 0.208]
  - model: OEvortex/Emotional-llama-8B
    parameters:
      density: 0.9
      gamma: 0.01
      weight: [0.139, 0.139, 0.208, 0.208, 0.139]
  - model: lighteternal/Llama3-merge-biomed-8b
    parameters:
      density: 0.9
      gamma: 0.01
      weight: [0.139, 0.208, 0.139, 0.208, 0.139]
  - model: Casual-Autopsy/Llama3-merge-psychotherapy-8b
    parameters:
      density: 0.9
      gamma: 0.01
      weight: [0.208, 0.139, 0.139, 0.139, 0.208]
merge_method: breadcrumbs_ties
base_model: TheSkullery/llama-3-cat-8b-instruct-v1
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-CRE


models:
  - model:  Casual-Autopsy/Super-Nova-CRE_pt.1
  - model: Casual-Autopsy/Super-Nova-CRE_pt.2
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-CRE_pt.1
parameters:
  t:
    - filter: self_attn
      value: [0.5, 0.3, 0.7, 0.5, 0.7, 0.3, 0.5, 0.3, 0.7, 0.5, 0.7, 0.3, 0.5]
    - filter: mlp
      value: [0.5, 0.7, 0.3, 0.5, 0.3, 0.7, 0.5, 0.7, 0.3, 0.5, 0.3, 0.7, 0.5]
    - value: 0.5
  embed_slerp: true
dtype: float32
out_dtype: bfloat16

Super-Nova-UNC

models:
  - model: Casual-Autopsy/Super-Nova-UNC_pt.1
  - model: Casual-Autopsy/Super-Nova-UNC_pt.2
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-UNC_pt.1
parameters:
  t:
    - value: [0.5, 0.7, 0.3, 0.5, 0.3, 0.7, 0.5, 0.7, 0.3, 0.5, 0.3, 0.7, 0.5]
  embed_slerp: true
dtype: float32
out_dtype: bfloat16

Super-Nova-INT

models:
  - model: Casual-Autopsy/Super-Nova-INT_pt.1
  - model: Casual-Autopsy/Super-Nova-INT_pt.2
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-INT_pt.1
parameters:
  t:
    - value: 0.5
  embed_slerp: true
dtype: float32
out_dtype: bfloat16

Super-Nova-RP_stp.1


models:
  - model: Casual-Autopsy/Super-Nova-CRE
  - model: asual-Autopsy/Super-Nova-UNC
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-CRE
parameters:
  t:
    - value: [0.7, 0.5, 0.3, 0.25, 0.2, 0.25, 0.3, 0.5, 0.7]
  embed_slerp: true
dtype: float32
out_dtype: bfloat16

Super-Nova-RP_stp.2

models:
  - model: Casual-Autopsy/Super-Nova-RP_stp.1
  - model: Casual-Autopsy/Super-Nova-INT
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-RP_stp.1
parameters:
  t:
    - value: [0.1, 0.15, 0.2, 0.4, 0.6, 0.4, 0.2, 0.15, 0.1]
  embed_slerp: true
dtype: float32
out_dtype: bfloat16

Super-Nova-RP_pt.1

models:
  - model: Casual-Autopsy/Super-Nova-RP_stp.2
  - model: Sao10K/L3-8B-Tamamo-v1
    parameters:
      density: [0.4, 0.6, 0.5, 0.6, 0.4]
      epsilon: [0.15, 0.15, 0.25, 0.15, 0.15]
      lambda: 0.85
      weight: [-0.01523, 0.01768, -0.01384, 0.01835, -0.01247]
  - model: ResplendentAI/Nymph_8B
    parameters:
      density: [0.65, 0.35, 0.5, 0.35, 0.65]
      epsilon: [0.1, 0.1, 0.25, 0.1, 0.1]
      lambda: 0.85
      weight: [0.01823, -0.01647, 0.01422, -0.01975, 0.01128]
  - model: ChaoticNeutrals/T-900-8B
    parameters:
      density: [0.35, 0.65, 0.5, 0.65, 0.35]
      epsilon: [0.1, 0.1, 0.25, 0.1, 0.1]
      lambda: 0.85
      weight: [-0.01891, 0.01554, -0.01325, 0.01791, -0.01458]
  - model: Sao10K/L3-8B-Niitama-v1
    parameters:
      density: [0.6, 0.4, 0.5, 0.4, 0.6]
      epsilon: [0.15, 0.15, 0.25, 0.15, 0.15]
      lambda: 0.85
      weight: [0.01768, -0.01675, 0.01285, -0.01696, 0.01421]
merge_method: della
base_model: Casual-Autopsy/Super-Nova-RP_stp.2
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

Super-Nova-RP_pt.2

models:
  - model: Casual-Autopsy/Super-Nova-RP_stp.2
  - model: bluuwhale/L3-SthenoMaidBlackroot-8B-V1
    parameters:
      density: [0.4, 0.6, 0.5, 0.6, 0.4]
      epsilon: [0.15, 0.15, 0.25, 0.15, 0.15]
      lambda: 0.85
      weight: [-0.01935, 0.01785, -0.01512, 0.01809, -0.01371]
  - model: Hastagaras/Jamet-8B-L3-MK.V-Blackroot
    parameters:
      density: [0.65, 0.35, 0.5, 0.35, 0.65]
      epsilon: [0.1, 0.1, 0.25, 0.1, 0.1]
      lambda: 0.85
      weight: [0.01847, -0.01468, 0.01503, -0.01822, 0.01459]
  - model: Hastagaras/Halu-8B-Llama3-Blackroot
    parameters:
      density: [0.35, 0.65, 0.5, 0.65, 0.35]
      epsilon: [0.1, 0.1, 0.25, 0.1, 0.1]
      lambda: 0.85
      weight: [-0.01578, 0.01821, -0.01753, 0.01677, -0.01442]
  - model: crestf411/L3-8B-sunfall-v0.4-stheno-v3.2
    parameters:
      density: [0.6, 0.5, 0.5, 0.5, 0.6]
      epsilon: [0.15, 0.15, 0.25, 0.15, 0.15]
      lambda: 0.85
      weight: [0.01667, -0.01740, 0.01560, -0.01564, 0.01315]
merge_method: della
base_model: Casual-Autopsy/Super-Nova-RP_stp.2
parameters:
  normalize: false
  int8_mask: true
dtype: float32
out_dtype: bfloat16

L3-Super-Nova-RP-8B

models:
  - model: Casual-Autopsy/Super-Nova-RP_pt.1
  - model: Casual-Autopsy/Super-Nova-RP_pt.2
merge_method: slerp
base_model: Casual-Autopsy/Super-Nova-RP_pt.1
parameters:
  t:
    - value: 0.5
dtype: float32
out_dtype: bfloat16
Downloads last month
381
GGUF
Model size
8.03B params
Architecture
llama

2-bit

3-bit

4-bit

5-bit

6-bit

8-bit

Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Model tree for QuantFactory/L3-Super-Nova-RP-8B-GGUF