G2-DA-Nyxora-27b-V2

Now that the cute anime girl has your attention.

Creator: SteelSkull

About G2-DA-Nyxora-27b-V2:

Model Name Legend =
"G2 = Gemma 2"
"DA = Della"
"Nyxora = Tiger Base = cat = Nya"
"27b = its 27b"
"V2 = V2"

This model is the second experimental foray into 27b models. Feedback is welcome for further improvements.

G2-DA-Nyxora-27b-V2 combines multiple models' strengths to provide a versatile assistant for various tasks, including general use, storytelling, and roleplay (ERP & RP).

The Model seems to have fixed alot of the prudish issues of the last model.

The new Exprimental merge method Della builds upon DARE, It uses adaptive pruning based on parameter magnitudes. DELLA first ranks parameters in each row of delta parameters and assigns drop probabilities inversely proportional to their magnitudes. This allows it to retain more important changes while reducing interference. After pruning, it rescales the remaining parameters similar to DARE.

Quants:

Will add as found or im notified of their creation (If you know of one create a discussion!)

SteelQuants, My Personal Quants

G2-DA-Nyxora-27b-V2-Q5_K_M-GGUF

G2-DA-Nyxora-27b-V2-Q4_K_M-GGUF

Config:

MODEL_NAME = "G2-DA-Nyxora-27b-V2"
yaml_config = """
base_model: TheDrummer/Big-Tiger-Gemma-27B-v1
merge_method: della
dtype: bfloat16
models:
  - model: TheDrummer/Gemmasutra-Pro-27B-v1
  - model: migtissera/Tess-v2.5-Gemma-2-27B-alpha
"""

Template:

<start_of_turn>user
{{ if .System }}{{ .System }} {{ end }}{{ .Prompt }}<end_of_turn>
<start_of_turn>model
{{ .Response }}<end_of_turn>

I've had a few people ask about donations so here's a link: