OpenCrystal-8B-L3

This is a far more optimized version of Darkknight535/OpenCrystal-12B-L3 which uses the original 8B models (instead of the upscaled 12B models).

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

Merge Details

Merge Method

This model was merged using the TIES merge method using unsloth/llama-3-8b-Instruct as a base.

Models Merged

The following models were included in the merge:

• nothingiisreal/L3-8B-Celeste-V1.2
• Sao10K/L3-8B-Niitama-v1
• unsloth/llama-3-8b-Instruct

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: unsloth/llama-3-8b-Instruct
  - model: nothingiisreal/L3-8B-Celeste-V1.2 # Another RP / Storytelling Model
    parameters:
      density: 0.1
      weight: 0.1
  - model: Sao10K/L3-8B-Niitama-v1 # Another RP / Storytelling Model
    parameters:
      density: 0.9
      weight: 0.9
merge_method: ties
base_model: unsloth/llama-3-8b-Instruct # For Base Coherence and prompting
parameters:
  int8_mask: true
  rescale: true
  normalize: false
dtype: bfloat16