schonsense/Tess_VAR_Brain_70B

Tess_brain

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

Merge Details

Merge Method

This model was merged using two base models and their spatially "awared" LORA fused versions.

• Step1 - Turn 2 models into 4 models via LORA
• Step2 - Model Stock
• Step3 - nuslerp1
• Step4 - nuslerp2
• Step5 - nuslerp final against Model stock merge as base.

Configuration - SCHEMATIC

The following YAML configuration was used to produce this model:

merge_method: nuslerp

models:

  - model: "D:\\mergekit\\_My_YAMLS\\Tess_brain_mslerp2"
    parameters:
      weight: 1

  - model: "D:\\mergekit\\_My_YAMLS\\Tess_brain_mslerp1"
    parameters:
      weight: 1

base_model: "D:\\mergekit\\_My_YAMLS\\Tess_brain_stock"

parameters:
  normalize: false
  int8_mask: true

---
name: Tess_brain_mslerp1
merge_method: nuslerp

models:

  - model: TareksTesting/Tesseract-V0.2-LLaMa-70B
    parameters:
      weight: 1

  - model: "D:\\mergekit\\Tess_0_2_VAR_r128"
    parameters:
      weight: 1

parameters:
  normalize: false
  int8_mask: true

---
name: Tess_brain_mslerp2
merge_method: nuslerp

models:

  - model: "D:\\mergekit\\Tess_2_0_VAR_r128"
    parameters:
      weight: 1

  - model: TareksTesting/Tesseract-V2.0-LLaMa-70B
    parameters:
      weight: 1

parameters:
  normalize: false
  int8_mask: true

---
name: Tess_brain_stock
models:
  - model: "D:\\mergekit\\Tess_2_0_VAR_r128"
  - model: TareksTesting/Tesseract-V2.0-LLaMa-70B
  - model: TareksTesting/Tesseract-V0.2-LLaMa-70B
  - model: "D:\\mergekit\\Tess_0_2_VAR_r128"
base_model: meta-llama/Llama-3.3-70B-Instruct
merge_method: model_stock
---

dtype: float32
out_dtype: bfloat16
chat_template: llama3
tokenizer:
 source: union