hell0ks/Smoothie-Seed-OSS-36B-Instruct-GGUF

Smoothie-qwen applied version of ByteDance-Seed/Seed-OSS-36B-Instruct

imatrix is created with bartowski's dataset

Tools used

• dnotitia/smoothie-qwen version: git+c005f0b

• llama.cpp version: 6636 (d72f5f7b)

What is Smoothie-Qwen?

By analyzing and adjusting token weights particularly those associated with specific Unicode ranges it helps mitigate unintended biases toward certain languages while preserving the model’s core capabilities.

Notes

• Use --temp 1.1 --top-p 0.95 --jinja for llama.cpp
• --chat_template_kwargs "{\"thinking_budget\": 0}" disables thinking