powermove72/LLama-4b-amt-v0.3

merge1

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:

• merge
• bunnycore/Llama-3.2-3b-RP-Toxic-Fuse
• EpistemeAI/ReasoningCore-Llama-3.2-3B-r1-v1_2

Configuration

The following YAML configuration was used to produce this model:

merge_method: passthrough
dtype: bfloat16
tokenizer_source: union
#base_model: huihui-ai/Hermes-3-Llama-3.2-3B-abliterated

slices:
  - sources:
      - model: EpistemeAI/ReasoningCore-Llama-3.2-3B-r1-v1_2
        layer_range: [0, 16]
        parameters:
          weight: 0.65  # Transcendent: HyperZone 1 - Efficiency Nexus (low layers with entropy scaling)
  - sources:
      - model: bunnycore/Llama-3.2-3b-RP-Toxic-Fuse
        layer_range: [16, 27]
        parameters:
          weight: 0.6  # Hyperscaled for transcendent boost
  - sources:
      - model: merge  # Reference the Step 1 output directory/model
        layer_range: [26, 27]
        # tensors: ["self_attn"]
        parameters:
          weight: 0.75  # Optimized via simulated genius annealing
        # tensors: ["mlp"]
        # parameters:
          # weight: 0.95