model

Any feedback regarding the model its behaviour is very welcome. In V1 i experimented with a giraffe base but i noticed the giraffe version quickly lossed the long context ability.

So in V1.5 i switched back to a gradient base.

Merge Details

Merge Method

This model was merged using the breadcrumbs_ties merge method using \Llama-3-70B-Instruct-Gradient-262k as a base.

Models Merged

The following models were included in the merge:

• \Llama-3-Lumimaid-70B-v0.1-OAS
• \Smaug-Llama-3-70B-Instruct
• \Llama-3-70B-Instruct-abliterated-v3

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: \Llama-3-70B-Instruct-Gradient-262k
    parameters:
      weight: 0.40
      density: 0.90
      gamma: 0.01
  - model: \Llama-3-70B-Instruct-abliterated-v3
    parameters:
      weight: 0.20
      density: 0.90
      gamma: 0.01
  - model: \Smaug-Llama-3-70B-Instruct
    parameters:
      weight: 0.40
      density: 0.90
      gamma: 0.01
  - model: \Llama-3-Lumimaid-70B-v0.1-OAS
    parameters:
      weight: 0.20
      density: 0.90
      gamma: 0.01
merge_method: breadcrumbs_ties
base_model: \Llama-3-70B-Instruct-Gradient-262k
dtype: bfloat16