zerofata/L3.3-GeneticLemonade-Final-70B-4bpw-h6-exl2

Genetic Lemonade Final

Inspired to learn how to merge by the Nevoria series from SteelSkull.

This model is the final result of the Genetic Lemonade series.

Designed for RP, this model is uncensored and focused around striking a balance between writing style, creativity and intelligence.

When compared to the previous Genetic Lemonade Unleashed, the final version:

• Better instruction following
• Less censored
• Slight intelligence drop
• Similar creativity
• Similar writing styles

The above is based on my testing with the model and benchmark scores from UGI leaderboard. Both models are relatively uncensored, but this one moreso than 'Unleashed'.

SillyTavern Settings

Llam@ception recommended for sane defaults if unsure, import them to SillyTavern and they're plug n play.

Sampler Settings

• Temp: 0.9-1.0
• MinP: 0.03-0.05
• Dry: 0.8, 1.75, 4

Temperature last, neutralize other samplers. This model natively strikes a balance of creativity & intelligence.

Instruct

Llama-3-Instruct-Names but you will need to uncheck "System same as user".

Quants

GGUF

• Static quants by mradermacher
• iMatrix quants by mradermacher

EXL2

• 4bpw
• 4.5bpw
• 6bpw

Merge Details

Merge Method

This model was merged using the SCE merge method.

The base aims to build a strong general purpose model using high performing models that are trained on various datasets from different languages / cultures. This is to reduce the chance of the same datasets appearing multiple times to build natural creativity into L3.3 The second merge aims to impart specific RP / creative writing knowledge, again focusing on trying to find high performing models that use or likely use different datasets.

Base_6_v2

models:
  - model: OpenBuddy/openbuddy-llama3.3-70b-v24.1-131k
  - model: nbeerbower/llama3.1-kartoffeldes-70B
  - model: tokyotech-llm/Llama-3.1-Swallow-70B-Instruct-v0.3
  - model: SicariusSicariiStuff/Negative_LLAMA_70B
select_topk: .15
merge_method: sce
base_model: meta-llama/Llama-3.3-70B-Instruct
out_dtype: bfloat16
dype: float32
tokenizer:
  source: base

Genetic Lemonade Final

models:
  - model: EVA-UNIT-01/EVA-LLaMA-3.33-70B-v0.1
  - model: LatitudeGames/Wayfarer-Large-70B-Llama-3.3
  - model: crestf411/L3.1-nemotron-sunfall-v0.7.0
  - model: SicariusSicariiStuff/Negative_LLAMA_70B
  - model: Sao10K/L3.3-70B-Euryale-v2.3
merge_method: sce
base_model: ./Base_6_v2
select_topk: 0.15
out_dtype: bfloat16
dype: float32
tokenizer:
  source: union