MrRobotoAI/MrRoboto-ProLong-8b-v4i

merge

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

Merge Details

Merge Method

This model was merged using the linear DARE merge method using MrRobotoAI/MrRoboto-ProLong-8b-v1n as a base.

Models Merged

The following models were included in the merge:

• MrRobotoAI/MrRoboto-ProLong-8b-v4h + Azazelle/Llama-3-LongStory-LORA
• MrRobotoAI/MrRoboto-ProLong-8b-v4b

Configuration

The following YAML configuration was used to produce this model:

merge_method: dare_linear
models:
  - model: MrRobotoAI/MrRoboto-ProLong-8b-v4b
    parameters:
      weight:
        - filter: v_proj
          value: [0.8, 0.8, 0.75, 0.55, 0.35, 0.15, 0.35, 0.55, 0.75, 0.8, 0.8]
        - filter: o_proj
          value: [0.8, 0.8, 0.75, 0.55, 0.35, 0.15, 0.35, 0.55, 0.75, 0.8, 0.8]
        - filter: up_proj
          value: [0.8, 0.8, 0.75, 0.55, 0.35, 0.15, 0.35, 0.55, 0.75, 0.8, 0.8]
        - filter: gate_proj
          value: [0.8, 0.8, 0.75, 0.55, 0.35, 0.15, 0.35, 0.55, 0.75, 0.8, 0.8]
        - filter: down_proj
          value: [0.8, 0.8, 0.75, 0.55, 0.35, 0.15, 0.35, 0.55, 0.75, 0.8, 0.8]
        - value: 1
  - model: MrRobotoAI/MrRoboto-ProLong-8b-v4h+Azazelle/Llama-3-LongStory-LORA
    parameters:
      weight:
        - filter: v_proj
          value: [0.2, 0.2, 0.25, 0.45, 0.65, 0.85, 0.65, 0.45, 0.25, 0.2, 0.2]
        - filter: o_proj
          value: [0.2, 0.2, 0.25, 0.45, 0.65, 0.85, 0.65, 0.45, 0.25, 0.2, 0.2]
        - filter: up_proj
          value: [0.2, 0.2, 0.25, 0.45, 0.65, 0.85, 0.65, 0.45, 0.25, 0.2, 0.2]
        - filter: gate_proj
          value: [0.2, 0.2, 0.25, 0.45, 0.65, 0.85, 0.65, 0.45, 0.25, 0.2, 0.2]
        - filter: down_proj
          value: [0.2, 0.2, 0.25, 0.45, 0.65, 0.85, 0.65, 0.45, 0.25, 0.2, 0.2]
        - value: 0
base_model: MrRobotoAI/MrRoboto-ProLong-8b-v1n
tokenizer_source: base
dtype: bfloat16