That merge might be worth something. Based on an experimental evolution of Smarteaze and the superb Electra R1 of Steelskull, as well as a 3.1 sidekick mixing brains and vanilla RP. I'm not fully satisfied yet, but we are getting somewhere. I'm still experimenting.

benchs

PPL 512 Wikitext Eng : 3.18
ARC-C : 60.55
ARC-E : 82.80

merge

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

Merge Details

Merge Method

This model was merged using the Model Stock merge method using NexesMess/Llama_3.x_70b_Tessessence_0.10c as a base.

Models Merged

The following models were included in the merge:

Configuration

The following YAML configuration was used to produce this model:

yaml

merge_method: model_stock
models:
  - model: Nexesenex/Llama_3.1_70b_Hostess_V1
    parameters:
      weight: 1.0
  - model: Steelskull/L3.3-Electra-R1-70b
    parameters:
      weight: 1.0
base_model: NexesMess/Llama_3.x_70b_Tessessence_0.10c
dtype: bfloat16
out_dtype: bfloat16
parameters:
  int8_mask: true
  normalize: true
  rescale: false
chat_template: auto
tokenizer:
  source: union

Nexesenex/Llama_3.x_70b_Triads_V3

about