dnotitia/smoothie-qwen
A lightweight adjustment tool for smoothing token probabilities in the Qwen models to encourage balanced multilingual generation.
This tool helps AI engineers and researchers working with Qwen language models to ensure more balanced outputs when generating text across multiple languages. It takes an existing Qwen model and a list of target Unicode ranges (representing specific languages or character sets) as input. The output is a modified Qwen model that generates text with a more even distribution of languages, preventing overrepresentation of certain languages.
104 stars. No commits in the last 6 months.
Use this if you are developing AI applications with Qwen models and need to prevent your model from over-generating text in certain languages, especially when aiming for diverse, multilingual outputs.
Not ideal if you are not working with Qwen models or if your application specifically requires an emphasis on a particular language rather than balanced multilingual generation.
Stars
104
Forks
7
Language
Python
License
—
Category
Last pushed
Jul 09, 2025
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/llm-tools/dnotitia/smoothie-qwen"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Higher-rated alternatives
huawei-csl/SINQ
Welcome to the official repository of SINQ! A novel, fast and high-quality quantization method...
SILX-LABS/QUASAR-SUBNET
QUASAR is a long-context foundation model and decentralized evaluation subnet built on Bittensor,
stackblogger/bitnet.js
BitNet.Js - A node.js implementation of the microsoft bitnet.cpp inference framework.
m96-chan/0xBitNet
Run BitNet b1.58 ternary LLMs with WebGPU — in browsers and native apps
AnswerDotAI/cold-compress
Cold Compress is a hackable, lightweight, and open-source toolkit for creating and benchmarking...