RichardErkhov/Gunulhona_-_Phi-Small-Merge-4bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

Phi-Small-Merge - bnb 4bits

• Model creator: https://huggingface.co/Gunulhona/
• Original model: https://huggingface.co/Gunulhona/Phi-Small-Merge/

Original model description:

base_model:

• Ejafa/phi-3-mini-128k-instruct-simpo-lr-5e-07-gamma-1.5
• jpacifico/Chocolatine-3B-Instruct-DPO-Revised
• Antonio88/TaliML-PHI3-128K-ITA-V.1.0.FINAL library_name: transformers tags:
• mergekit
• merge

---

merge

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

Merge Details

Merge Method

This model was merged using the SLERP merge method.

Models Merged

The following models were included in the merge:

• Ejafa/phi-3-mini-128k-instruct-simpo-lr-5e-07-gamma-1.5
• jpacifico/Chocolatine-3B-Instruct-DPO-Revised
• Antonio88/TaliML-PHI3-128K-ITA-V.1.0.FINAL

Configuration

The following YAML configuration was used to produce this model:

slices:
  - sources:
    - model: Antonio88/TaliML-PHI3-128K-ITA-V.1.0.FINAL
      layer_range: [0, 32]
  - sources:
    - model: jpacifico/Chocolatine-3B-Instruct-DPO-Revised
      layer_range: [0, 32]
  - sources:
    - model: Ejafa/phi-3-mini-128k-instruct-simpo-lr-5e-07-gamma-1.5
      layer_range: [0, 32]
base_model: jpacifico/Chocolatine-3B-Instruct-DPO-Revised
merge_method: slerp
parameters:
  t:
    - filter: self_attn
      value: [0, 0.5, 0.3, 0.7, 1]
    - filter: mlp
      value: [1, 0.5, 0.7, 0.3, 0]
    - value: 0.5
  normalize: false
  int8_mask: true
  density: 0.7
  lambda: 1.1
  epsilon: 0.2
dtype: bfloat16