Evolved-Llama3-8B

Evolved-Llama3-8B is a merge of the following models using mergekit:

• elyza/Llama-3-ELYZA-JP-8B
• nvidia/Llama3-ChatQA-1.5-8B

🧩 Configuration

slices:
- sources:
  - layer_range: [0, 8]
    model: Llama-3-ELYZA-JP-8B_2371007997
    parameters:
      weight: 0.2924041594566723
  - layer_range: [0, 8]
    model: Llama3-ChatQA-1.5-8B_376305873
    parameters:
      weight: 1.0002597402802504
- sources:
  - layer_range: [8, 16]
    model: Llama-3-ELYZA-JP-8B_2371007997
    parameters:
      weight: 0.5303090111436538
  - layer_range: [8, 16]
    model: Llama3-ChatQA-1.5-8B_376305873
    parameters:
      weight: 0.6266010695928661
- sources:
  - layer_range: [16, 24]
    model: Llama-3-ELYZA-JP-8B_2371007997
    parameters:
      weight: 0.3491957124910876
  - layer_range: [16, 24]
    model: Llama3-ChatQA-1.5-8B_376305873
    parameters:
      weight: 0.44349113433925463
- sources:
  - layer_range: [24, 32]
    model: Llama-3-ELYZA-JP-8B_2371007997
    parameters:
      weight: 0.38380980665908515
  - layer_range: [24, 32]
    model: Llama3-ChatQA-1.5-8B_376305873
    parameters:
      weight: 0.5068229626895051