Llama-3-BioHacker-8B

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

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

• aaditya/Llama3-OpenBioLLM-8B
• WhiteRabbitNeo/Llama-3-WhiteRabbitNeo-8B-v2.0

Configuration

The following YAML configuration was used to produce this model:

# Bio-Hacker Chimera Merge Configuration
# SLERP merge of Biomedical + Cybersecurity Llama-3 models
#
# The parameter oscillation creates a gradient through layers:
# - Early layers: Bio-dominant attention, Cyber-dominant MLP
# - Middle layers: Balanced mix
# - Late layers: Cyber-dominant attention, Bio-dominant MLP
#
# This should create interesting cross-domain "thinking patterns"

slices:
  - sources:
      - model: aaditya/Llama3-OpenBioLLM-8B
        layer_range: [0, 32]
      - model: WhiteRabbitNeo/Llama-3-WhiteRabbitNeo-8B-v2.0
        layer_range: [0, 32]

merge_method: slerp

base_model: aaditya/Llama3-OpenBioLLM-8B

parameters:
  t:
    # Self-attention: Start Bio (0), end Cyber (1)
    # Controls "what to pay attention to"
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    # MLP: Start Cyber (1), end Bio (0)
    # Controls "how to transform information"
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    # Default for anything else (embeddings, norms, etc)
    - value: 0.5

dtype: float16