Modelpublic

wvnvwn/llama2-7b-chat-lr5e-5-winogrande-lr5e-5-resta0_3

sourceHugging Faceupdated 2mo agoView on Hugging Face

0likes14downloads

Model Card

winogrande

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

Merge Details

Merge Method

This model was merged using the linear merge method.

Models Merged

The following models were included in the merge:

—/NHNHOME/26msit001A/BASE/edgeailab/jongbokwon/warp-safety/outputs/chatssft/llama27bchatcircuitbreakers
—/NHNHOME/26msit001A/BASE/edgeailab/jongbokwon/minseongresults/checkpoints/llama-2-7b-chat-hf/baselines/vanilla/winogrande
—/NHNHOME/26msit001A/BASE/edgeailab/jongbokwon/warp-safety/outputs/base/llama27bchathf

Configuration

The following YAML configuration was used to produce this model:

yaml

dtype: float16
merge_method: linear
slices:
- sources:
  - layer_range: [0, 32]
    model:
      model:
        path: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/minseong_results/checkpoints/llama-2-7b-chat-hf/baselines/vanilla/winogrande
    parameters:
      weight: 1.0
  - layer_range: [0, 32]
    model:
      model:
        path: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/warp-safety/outputs/chat_ssft/llama2_7b_chat_circuit_breakers
    parameters:
      weight: 0.3
  - layer_range: [0, 32]
    model:
      model:
        path: /NHNHOME/26msit001_A/BASE/edge_ai_lab/jongbokwon/warp-safety/outputs/base/llama2_7b_chat_hf
    parameters:
      weight: -0.3