Muhammad2003/TriMistral-7B-SLERP

merge

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:

• HuggingFaceH4/zephyr-7b-beta
• NousResearch/Hermes-2-Pro-Mistral-7B
• instructlab/merlinite-7b-lab

Configuration

Since Slerp allows merging two models at a time, the following YAML configurations were used to produce this model:

slices:
  - sources:
      - model: HuggingFaceH4/zephyr-7b-beta
        layer_range: [0, 32]
      - model: NousResearch/Hermes-2-Pro-Mistral-7B
        layer_range: [0, 32]
merge_method: slerp
base_model: HuggingFaceH4/zephyr-7b-beta
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

Then

slices:
  - sources:
      - model: ./merge
        layer_range: [0, 32]
      - model: instructlab/merlinite-7b-lab
        layer_range: [0, 32]
merge_method: slerp
base_model: ./merge
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
dtype: bfloat16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	67.76
AI2 Reasoning Challenge (25-Shot)	64.25
HellaSwag (10-Shot)	85.47
MMLU (5-Shot)	64.89
TruthfulQA (0-shot)	53.57
Winogrande (5-shot)	79.16
GSM8k (5-shot)	59.21