Downloads · 30 days
0
leeminwaan/nekochan-molo32-qwen3-1.7B
nekochan-molo32-qwen3-1.7B is a text generation model from leeminwaan. Use it when you need the model to write or continue text. It is set up for transformers. The card lists the license as apache-2.0.
Downloads · 30 days
0
Access
Public
Updated Dec 11, 2025
Repo size
848 MB
Likes
0
Public
Click a slice to open those files.
.safetensors837 MB · 98%
From the Hugging Face model README
A MoLo-enhanced variant of Qwen3-1.7B.
Nekochan-Molo32-Qwen3-1.7B is a custom model built on top of Qwen3-1.7B-Base, augmented with a MoLo (Mixture-of-LoRA-Experts) architecture (My idea). This model blends Qwen’s strong base capabilities with a lightweight expert-routing system, allowing it to adopt different style “modes” depending on the input.
This release focuses on:
⚠️ Note:
The full architecture implementation (MoLoModel.py and MoLo layers) will be published soon (maybe) and later packaged into a dedicated library for easy import.
This model is primarily designed for text generation with unique stylistic blending enabled through the MoLo expert routing system.
Nope, as it requires a custom model structure..
Nekochan-Molo32-Qwen3-1.7B can be served through any framework that supports Hugging Face-style causal language models, including:
Once the MoLo architecture library is released, these platforms will also support full MoLo expert routing inference without additional configuration.
To get the best results from Nekochan-Molo32-Qwen3-1.7B:
Sampling Recommendations
Long-Context Usage
Style Control via Experts
Router Stability
The MoLo architecture source code:
MoLoModel.pyMoLoLinearwill be uploaded soon and later bundled into a dedicated pip installable library for simple usage:
pip install molo-neko
Once published, users will be able to load the model like:
from molo_neko import MoLoModel
model = MoLoModel.from_pretrained("leeminwaan/Nekochan-Molo32-Qwen3-1.7B")
Stay tuned!
PS: I'll train the model first then release the library later, just to test the efficiency PS-1.1: Only the adapter will be pushed.