Magnolia-v5a-12B

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

Again Nemo 12B Base was used as a base for task arithmetic merger, rather than 12B Instruct or a variant. A model fine-tuned with mostly Instruct and some narrative generation data was added, along with an abliterated Instruct model,which curiously benched higher than Nemo 12B Instruct itself.

Merge Details

Merge Method

This model was merged using the Task Arithmetic merge method using grimjim/mistralai-Mistral-Nemo-Base-2407 as a base.

Models Merged

The following models were included in the merge:

• grimjim/Magnolia-v3-12B
• Nitral-AI/Captain_BMO-12B
• natong19/Mistral-Nemo-Instruct-2407-abliterated

Configuration

The following YAML configuration was used to produce this model:

base_model: grimjim/mistralai-Mistral-Nemo-Base-2407
dtype: bfloat16
merge_method: task_arithmetic
parameters:
  normalize: true
models:
  - model: grimjim/mistralai-Mistral-Nemo-Base-2407
  - model: grimjim/Magnolia-v3-12B
    parameters:
      weight: 0.85
  - model: Nitral-AI/Captain_BMO-12B
    parameters:
      weight: 0.1
  - model: natong19/Mistral-Nemo-Instruct-2407-abliterated
    parameters:
      weight: 0.05