Nexesenex/Llama_3.x_70b_Hexagon_Blue_V1

about

Smatricks_v1.30_flat + DarkHorse for a darker colloration.

I started to use it, and I'm satisfied.

• In the line of the Smartricks v1.30 flat, its prose is quite different of what I usually observe in my merges based on a smart merge.
• The addition of DarkHorse brings more creativity without damaging much the smarts.
• I continue to think that 2 levels of merges is quite optimal for a final model when using merge_stock. Beyond, it becomes more "soupy".

---

benchs

IK_LLama.CPP Benchs in IQ6_K:

• PPL-512 WikiEng Text 564 : 3.40
• ARC-C 299 : 59.87
• ARC-E 570 : 81.23
• Hellaswag 200 : 86.5
• Winogrande 1263 : 81.92
• MMLU : 46.90

MMLU results are much lower than they should be on LlamaCPP. This is a constant quirk since 2024.

---

merge

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

Merge Details

Merge Method

This model was merged using the Model Stock merge method using huihui-ai/Llama-3.3-70B-Instruct-abliterated as a base.

Models Merged

The following models were included in the merge:

• huihui-ai/Llama-3.1-Nemotron-70B-Instruct-HF-abliterated
• TheDrummer/Fallen-Llama-3.3-R1-70B-v1
• huihui-ai/Llama-3.1-Tulu-3-70B-abliterated
• Nexesenex/Llama_3.3_70b_DarkHorse
• hitachi-nlp/Llama-3.1-70B-FLDx2

Configuration

The following YAML configuration was used to produce this model:

merge_method: model_stock
models:
  - model: TheDrummer/Fallen-Llama-3.3-R1-70B-v1
    parameters:
      weight: 1.0
  - model: Nexesenex/Llama_3.3_70b_DarkHorse
    parameters:
      weight: 1.0
  - model: huihui-ai/Llama-3.1-Nemotron-70B-Instruct-HF-abliterated
    parameters:
      weight: 1.0
  - model: huihui-ai/Llama-3.1-Tulu-3-70B-abliterated
    parameters:
      weight: 1.0
  - model: hitachi-nlp/Llama-3.1-70B-FLDx2
    parameters:
      weight: 1.0
base_model: huihui-ai/Llama-3.3-70B-Instruct-abliterated
dtype: bfloat16
out_dtype: bfloat16
parameters:
  int8_mask: true
  normalize: true
  rescale: false
  filter_wise: false
  smooth: false
  allow_negative_weights: false
chat_template: auto
tokenizer:
  source: union