DreadPoor/WIP_Damascus-8B-TIES

merge

OI, if this merge was a game, it would be as good as CONCORD was.

that really was the biggest insult I could think of, which says some about me, but enough about this merge.

Simply put, don't bother with this one. It was a gamble, and it didn't pan out.

Merge Details

Merge Method

This model was merged using the TIES merge method using NousResearch/Meta-Llama-3-8B + grimjim/Llama-3-Instruct-abliteration-LoRA-8B as a base.

Models Merged

The following models were included in the merge:

• DreadPoor/Aurora_faustus-8B-LINEAR
• DreadPoor/Aspire1.2-8B-TIES
• DreadPoor/Heart_Stolen-8B-Model_Stock

Configuration

The following YAML configuration was used to produce this model:

slices:
- sources:
  - layer_range: [0, 32]
    model: DreadPoor/Heart_Stolen-8B-Model_Stock
    parameters:
      weight: 1
  - layer_range: [0, 32]
    model: DreadPoor/Aurora_faustus-8B-LINEAR
    parameters:
      weight: 1
  - layer_range: [0, 32]
    model: DreadPoor/Aspire1.2-8B-TIES
    parameters:
      weight: 1
  - layer_range: [0, 32]
    model: NousResearch/Meta-Llama-3-8B+grimjim/Llama-3-Instruct-abliteration-LoRA-8B
    parameters:
      density: 1

base_model: NousResearch/Meta-Llama-3-8B+grimjim/Llama-3-Instruct-abliteration-LoRA-8B
merge_method: ties
dtype: bfloat16
parameters:
  normalize: true
  int8_mask: true

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	24.73
IFEval (0-Shot)	47.76
BBH (3-Shot)	34.52
MATH Lvl 5 (4-Shot)	15.11
GPQA (0-shot)	7.61
MuSR (0-shot)	12.72
MMLU-PRO (5-shot)	30.68