llama-3-sqrt-crocodile-v0.2A

This is a merge of pre-trained language models created using mergekit.
Merge Details

Merge Method

This model was merged using the passthrough merge method.
Models Merged

The following models were included in the merge:
llama-3-sqrt-crocodile-v0.0A/the-operator
llama-3-sqrt-crocodile-v0.0A/sqrt-talker
Configuration

The following YAML configuration was used to produce this model:
models:
  - model: Orenguteng/Lexi-Llama-3-8B-Uncensored
    parameters:
      weight: [0.2, 0.3, 0.4, 0.6]
    layer_range: [0, 32]
  - model: NousResearch/Meta-Llama-3-8B
    parameters:
      weight: [0.6, 0.2, 0.2, 0.1]
    layer_range: [0, 32]
  - model: NousResearch/Meta-Llama-3-8B-Instruct
    parameters:
      weight: [0.2, 0.3, 0.85, 0.3]
    layer_range: [0, 32]
merge_method: dare_linear
base_model: NousResearch/Meta-Llama-3-8B-Instruct
dtype: bfloat16
name: Uninstruct-Uncensored
---
models:
  - model: cognitivecomputations/dolphin-2.9-llama3-8b
    parameters:
      weight: [0.25, 0.4, 0.35, 0.35]
      density: [0.3, 0.45, 0.2, 0.6]
    layer_range: [0, 32]
  - model: NousResearch/Meta-Llama-3-8B
    parameters: 
      weight: [0.15, 0.25, 0.05, 0]
      density: [0.2, 0.3, 0.4, 0.1]
  - model: Undi95/Llama-3-Unholy-8B
    parameters:
      weight: [0.4, 0.25, 0.45, 0.35]
      density: [0.2, 0.15, 1.5, 0.1]
    layer_range: [0, 32]
  - model: Uninstruct-Uncensored
    parameters:
      weight: [0.3, 0.1, 0.25, 0.3]
      density: [0.3, 0.15, 2.5, 0.2]
    layer_range: [0, 32]
merge_method: dare_ties
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: augmented-dolphin-hap
---
models:
  - model: vicgalle/Configurable-Llama-3-8B-v0.3
    parameters:
      weight: [0.5, 0.3, 0.1]
  - model: hiieu/Meta-Llama-3-8B-Instruct-function-calling-json-mode
    parameters:
      weight: 0.5
  - model: Trelis/Meta-Llama-3-8B-Instruct-function-calling
    parameters:
      weight: 0.3
    layer_range: [0, 32]
  - model: Rookie/Llama-3-8B-Instruct-Chinese 
    parameters:
      weight: 0.2
    layer_range: [0, 32]
  - model: Uninstruct-Uncensored
    parameters:
      weight: [0.7, 0.4, 0.25, 0.1]
    layer_range: [0, 32]
merge_method: model_stock
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: the-operator
---
models:
  - model: vicgalle/Configurable-Llama-3-8B-v0.3
    parameters:
      weight: 0.7
  - model: hiieu/Meta-Llama-3-8B-Instruct-function-calling-json-mode
    parameters:
      weight: 0.1
  - model: Trelis/Meta-Llama-3-8B-Instruct-function-calling
    parameters:
      weight: 0.03
    layer_range: [0, 32]
  - model: Rookie/Llama-3-8B-Instruct-Chinese
    parameters:
      weight: 0.07
    layer_range: [0, 32]
  - model: Uninstruct-Uncensored
    parameters:
      weight: 0.1
    layer_range: [0, 32]
merge_method: model_stock
base_model: Uninstruct-Uncensored
dtype: bfloat16
name: her-calculator
---
models:
  - model: her-calculator
    parameters:
      density: 0.7 # density gradient
      weight: [0.7, 0.5, 0.1, 0.8]
  - model: augmented-dolphin-hap
    parameters:
      weight: 0.7
merge_method: slerp
base_model: her-calculator
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5 # fallback for rest of tensors
dtype: float16
name: sqrt-talker
---
slices:
  - sources:
    - model: the-operator 
      layer_range: [0, 6]
  - sources:
    - model: sqrt-talker
      layer_range: [3, 9]
  - sources:
    - model: the-operator 
      layer_range: [6, 12]
  - sources:
    - model: sqrt-talker
      layer_range: [10, 16]
  - sources:
    - model: the-operator 
      layer_range: [13, 19]
  - sources:
    - model: sqrt-talker
      layer_range: [16, 22]
  - sources:
    - model: the-operator 
      layer_range: [19, 25]
  - sources:
    - model: sqrt-talker
      layer_range: [22, 28]
  - sources:
    - model: the-operator 
      layer_range: [26, 32]
merge_method: passthrough
dtype: bfloat16
name: llama-3-sqrt-crocodile-v0.2A