Llama-3.1-8B-InitializedEmbeddings_with_Hermes-3

Llama-3.1-8B-InitializedEmbeddings_with_Hermes-3

This is Meta's LLaMA-3.1-8B (base model) pre-initialized to NousResearch/Hermes-3-Llama-3.1-8B's embedding's special tokens (prompt/chat template).

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

Credit for the idea of this merge goes to Charles Goddard (Creator of mergekit); his merge chargoddard/Meta-Llama-3-8B-InitializedEmbeds outlined the details for this merge and explained how it all works, and why it is necessary to pre-initialize the base model with the instruct model's embedding's special tokens.

Merge Details

Merge Method

This model was merged using the linear merge method.

Models Merged

The following models were included in the merge:

• /Users/jsarnecki/opt/Workspace/meta-llama/Llama-3.1-8B
• /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B

Configuration

The following YAML configuration was used to produce this model:

merge_method: linear
dtype: float32
out_dtype: bfloat16
models:
  - model: /Users/jsarnecki/opt/Workspace/meta-llama/Llama-3.1-8B
    parameters:
      weight: 1.0
  - model: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B 
    parameters:
      weight: 0.0
tokenizer:
  source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
  tokens:
    <|start_header_id|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|end_header_id|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|eot_id|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|begin_of_text|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|end_of_text|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|im_start|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true
    <|im_end|>:
      source: /Users/jsarnecki/opt/Workspace/NousResearch/Hermes-3-Llama-3.1-8B
      force: true