6bpw exl2 quant of: https://huggingface.co/Nohobby/MS-Schisandra-22B-v0.2

Schisandra

Many thanks to the authors of the models used!

Overview

Main uses: RP, Storywriting

Prompt format: Mistral-V3

An intelligent model that is attentive to details and has a low-slop writing style. This time with a stable tokenizer.

Oh, and it now contains 10 finetunes! Not sure if some of them actually contribute to the output, but it's nice to see the numbers growing.

Quants

GGUF: Static | Imatrix

exl2: 4.65bpw 5.5bpw 6.5bpw

Settings

My SillyTavern preset: https://huggingface.co/Nohobby/MS-Schisandra-22B-v0.2/resolve/main/ST-formatting-Schisandra.json

Merge Details

Merging steps

Step1

(Config partially taken from here)

base_model: spow12/ChatWaifu_v2.0_22B
parameters:
  int8_mask: true
  rescale: true
  normalize: false
dtype: bfloat16
tokenizer_source: base
merge_method: della
models:
  - model: Envoid/Mistral-Small-NovusKyver
    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: rAIfle/Acolyte-22B
    parameters:
      density: [0.6, 0.4, 0.5, 0.4, 0.6]
      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]

Step2

(Config partially taken from here)

base_model: InferenceIllusionist/SorcererLM-22B
parameters:
  int8_mask: true
  rescale: true
  normalize: false
dtype: bfloat16
tokenizer_source: base
merge_method: della
models:
  - model: crestf411/MS-sunfall-v0.7.0
    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: anthracite-org/magnum-v4-22b
    parameters:
      density: [0.6, 0.4, 0.5, 0.4, 0.6]
      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]

SchisandraVA2

(Config taken from here)

merge_method: della_linear
dtype: bfloat16
parameters:
  normalize: true
  int8_mask: true
tokenizer_source: base
base_model: TheDrummer/UnslopSmall-22B-v1
models:
    - model: ArliAI/Mistral-Small-22B-ArliAI-RPMax-v1.1
      parameters:
        density: 0.55
        weight: 1
    - model: Gryphe/Pantheon-RP-Pure-1.6.2-22b-Small
      parameters:
        density: 0.55
        weight: 1
    - model: Step1
      parameters:
        density: 0.55
        weight: 1
    - model: allura-org/MS-Meadowlark-22B
      parameters:
        density: 0.55
        weight: 1
    - model: Step2
      parameters:
        density: 0.55
        weight: 1

Schisandra-v0.2

dtype: bfloat16
tokenizer_source: base
merge_method: della_linear
parameters:
  density: 0.5
base_model: SchisandraVA2
models:
  - model: unsloth/Mistral-Small-Instruct-2409
    parameters:
      weight:
        - filter: v_proj
          value: [0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0]
        - filter: o_proj
          value: [1, 0, 1, 0, 0, 0, 0, 0, 1, 1, 1]
        - filter: up_proj
          value: [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
        - filter: gate_proj
          value: [0, 0, 1, 1, 1, 1, 1, 1, 1, 0, 0]
        - filter: down_proj
          value: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
        - value: 0
  - model: SchisandraVA2
    parameters:
      weight:
        - filter: v_proj
          value: [1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1]
        - filter: o_proj
          value: [0, 1, 0, 1, 1, 1, 1, 1, 0, 0, 0]
        - filter: up_proj
          value: [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
        - filter: gate_proj
          value: [1, 1, 0, 0, 0, 0, 0, 0, 0, 1, 1]
        - filter: down_proj
          value: [1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1]
        - value: 1

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	30.22
IFEval (0-Shot)	63.83
BBH (3-Shot)	40.61
MATH Lvl 5 (4-Shot)	19.94
GPQA (0-shot)	11.41
MuSR (0-shot)	10.67
MMLU-PRO (5-shot)	34.85

Jellon
/

MS-Schisandra-22B-v0.2-exl2-6bpw-h8