Bragi3

Too sloppy for my tastes.

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

Merge Details

Merge Method

This model was merged using the NuSLERP merge method using meta-llama/Llama-3.1-70B as a base.

Models Merged

The following models were included in the merge:

• CrucibleLab/L3.3-70B-Loki-V2.0
• schonsense/Tropoplectic

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: CrucibleLab/L3.3-70B-Loki-V2.0
    parameters:        
          weight:
            - filter: q_proj
              value: [0.80, 0.30, 0.30, 0.30, 0.8]
            - filter: k_proj
              value: [0.70, 0.20, 0.20, 0.20, 0.7]
            - filter: v_proj
              value: [0.80, 0.40, 0.40, 0.40, 0.8]
            - filter: o_proj
              value: [0.90, 0.80, 0.80, 0.80, 0.9] 
            - filter: gate_proj
              value: [0.80, 0.20, 0.20, 0.20, 0.8]
            - filter: up_proj
              value: [0.80, 0.30, 0.30, 0.30, 0.8]
            - filter: down_proj
              value: [0.90, 0.80, 0.80, 0.80, 0.9]
            - filter: lm_head
              value: 0.95
            - value: 1



  - model: schonsense/Tropoplectic
    parameters:       
          weight:
            - filter: q_proj
              value: [0.20, 0.70, 0.70, 0.70, 0.2]
            - filter: k_proj
              value: [0.30, 0.80, 0.80, 0.80, 0.3]
            - filter: v_proj
              value: [0.20, 0.60, 0.60, 0.60, 0.2]
            - filter: o_proj
              value: [0.10, 0.25, 0.25, 0.25, 0.1] 
            - filter: gate_proj
              value: [0.20, 0.80, 0.80, 0.80, 0.2]
            - filter: up_proj
              value: [0.20, 0.70, 0.70, 0.70, 0.2]
            - filter: down_proj
              value: [0.10, 0.25, 0.25, 0.25, 0.1]
            - filter: lm_head
              value: 0.05
            - value: 0

base_model: meta-llama/Llama-3.1-70B
merge_method: nuslerp

parameters:
  normalize: false
  int8_mask: false
  rescale: false

dtype: float32
out_dtype: bfloat16

chat_template: llama3
tokenizer:
  source: union
  pad_to_multiple_of: 8