EurobeatVARemix-Qwen2.5-72b

Updated EVA to 0.1. That's all folks!

...It didn't feel right calling it LLENN anymore so I'm changing the name. "Pray I don't alter it any further."

Please do not ask for quants, contact others instead.

All models are ready for testing on featherless.ai as soon as it goes live.

Merge Details

Merge Method

This model was merged using the Model Stock merge method using Qwen/Qwen2.5-72B as a base.

Prompt Format

ChatML works for the most part.

Sampler Settings

Personally I use the following:

Temp: 1.2
Min P: 0.07
Rep Pen: 1.1

Others have suggested the following:

Temp: 1.1
Top P: 0.98
Min P: 0.05

Models Merged

The following models were included in the merge:

• ZeusLabs/Chronos-Platinum-72B
• EVA-UNIT-01/EVA-Qwen2.5-72B-v0.1
• m8than/banana-2-b-72b
• abacusai/Dracarys2-72B-Instruct
• rombodawg/Rombos-LLM-V2.5-Qwen-72b

Configuration

The following YAML configuration was used to produce this model:

models:
  - model: EVA-UNIT-01/EVA-Qwen2.5-72B-v0.1
  - model: ZeusLabs/Chronos-Platinum-72B
  - model: abacusai/Dracarys2-72B-Instruct
  - model: rombodawg/Rombos-LLM-V2.5-Qwen-72b
  - model: m8than/banana-2-b-72b

merge_method: model_stock
base_model: Qwen/Qwen2.5-72B
parameters:
  normalize: true
dtype: bfloat16