README.md

metadata

base_model: []
library_name: transformers
tags:
  - mergekit
  - merge

FuseChat-7B-Ties-VaRM-comb1-top0.1

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

Merge Details

Merge Method

This model was merged using the ties_varm merge method using /data/wanfq/fusechat/models/openchat_3.5 as a base.

Models Merged

The following models were included in the merge:

• /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Mixtral
• /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-NHSolar
• /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Qwen
• /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-internlm
• /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Starling

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: /data/wanfq/fusechat/models/openchat_3.5
    # no parameters necessary for base model
  - model: /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-internlm
    parameters:
      density: 0.1 # density gradient
      weight: 0.2
  - model: /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Qwen
    parameters:
      density: 0.5
      weight: 0.2 # weight gradient
  - model: /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Mixtral
    parameters:
      density: 0.5
      weight: 0.2 # weight gradient
  - model: /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-Starling
    parameters:
      density: 0.5
      weight: 0.2 # weight gradient
  - model: /data/wanfq/agno_fusion/fusechat_models/OpenChat-3.5-7B-NHSolar
    parameters:
      density: 0.5
      weight: 0.2 # weight gradient
merge_method: ties_varm
base_model: /data/wanfq/fusechat/models/openchat_3.5
parameters:
  normalize: true
dtype: bfloat16