saishf/Multi-Verse-RP-7B

Multi verse img!

merge

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

Merge Details

• This merge is entirely experimental, I've only tested it a few times but it seems to work? Thanks for all the loras jeiku. I keep getting driver crashes training my own :\
• Update, It scores well! My highest scoring model so far
• Self testing results, it can handle non-human characters surprisingly well and does well seperating human actions from non-human actions. I'm happy with it :3
• Works with alpaca best, Loras' are alpaca. But works with chatml too!

Merge Method

This model was merged using the task arithmetic merge method using ammarali32/multi_verse_model as a base.

Models Merged

The following models were included in the merge:

• ammarali32/multi_verse_model + jeiku/Theory_of_Mind_Roleplay_Mistral
• ammarali32/multi_verse_model + jeiku/Alpaca_NSFW_Shuffled_Mistral
• ammarali32/multi_verse_model + jeiku/Theory_of_Mind_Mistral
• ammarali32/multi_verse_model + jeiku/Gnosis_Reformatted_Mistral
• ammarali32/multi_verse_model + jeiku/Re-Host_Limarp_Mistral
• ammarali32/multi_verse_model + jeiku/Luna_LoRA_Mistral

Configuration

The following YAML configuration was used to produce this model:

merge_method: task_arithmetic
base_model: ammarali32/multi_verse_model
parameters:
  normalize: true
models:
  - model: ammarali32/multi_verse_model+jeiku/Gnosis_Reformatted_Mistral
    parameters:
      weight: 0.7
  - model: ammarali32/multi_verse_model+jeiku/Theory_of_Mind_Roleplay_Mistral
    parameters:
      weight: 0.65
  - model: ammarali32/multi_verse_model+jeiku/Luna_LoRA_Mistral
    parameters:
      weight: 0.5
  - model: ammarali32/multi_verse_model+jeiku/Re-Host_Limarp_Mistral
    parameters:
      weight: 0.8
  - model: ammarali32/multi_verse_model+jeiku/Alpaca_NSFW_Shuffled_Mistral
    parameters:
      weight: 0.75  
  - model: ammarali32/multi_verse_model+jeiku/Theory_of_Mind_Mistral
    parameters:
      weight: 0.7                     
dtype: float16

Open LLM Leaderboard Evaluation Results

Detailed results can be found here

Metric	Value
Avg.	74.73
AI2 Reasoning Challenge (25-Shot)	72.35
HellaSwag (10-Shot)	88.37
MMLU (5-Shot)	63.94
TruthfulQA (0-shot)	73.19
Winogrande (5-shot)	84.14
GSM8k (5-shot)	66.41