RichardErkhov/berkecr_-_tr-dare-merge-7B-4bits

Quantization made by Richard Erkhov.

Github

Discord

Request more models

tr-dare-merge-7B - bnb 4bits

• Model creator: https://huggingface.co/berkecr/
• Original model: https://huggingface.co/berkecr/tr-dare-merge-7B/

Original model description:

license: apache-2.0 tags: - merge - mergekit - lazymergekit - Mistral-7B-Instruct-v0.2 language: - tr - en

tr-dare-merge-7B

Configs

models:
  - model: mistralai/Mistral-7B-Instruct-v0.2
  - model: TURKCELL/Turkcell-LLM-7b-v1
    parameters:
      density: 0.7
      weight: 0.4
  - model: Trendyol/Trendyol-LLM-7b-chat-dpo-v1.0
    parameters:
      density: 0.7
      weight: 0.4
merge_method: dare_ties
base_model: mistralai/Mistral-7B-Instruct-v0.2
parameters:
  int8_mask: true
dtype: bfloat16