base_model, tags

base_model	tags
Qwen/Qwen3-1.7B	peft lora text-generation

Qwen3-1.7B-seed_gen_voronoi

Model Description

Fine-tuned from Qwen/Qwen3-1.7B using QLoRA (4-bit) with supervised fine-tuning.

Training Details

• Dataset: Dannys0n/test-dataset
• LoRA rank: 16, alpha: 32
• Epochs: 3, Learning rate: 0.0002

Intended Use

This model is a test model used for the CS-394/594 class at DigiPen.

The model is designed to provide seed vectors as structured json for generating voronoi diagrams.
This can be used just by prompting the model with the example prompt for output, or you can run a local server in LMstudio and run this python notebook locally for generating voronoi visualizations as well https://github.com/dannys0n/CS394-MyModulesRepo/tree/main/notebooks/07

Limitations

This model is a single-turn model and has not been trained on support long, multi-turn conversations.

Example Prompt

You are placing shard seed centers in a normalized 2D world.

Coordinates are in [0,1] x [0,1]. The world aspect ratio is width/height.

Goal: Place K shard centers that balance load from hotspots.

Hotspots represent player density and include:

• x, y position
• weight (importance)
• radius (influence area)

Guidelines:

• Keep shard centers within [0,1] bounds
• Prefer placing centers near clusters of hotspots
• Avoid placing centers too close together
• Try to distribute centers so hotspot load is balanced

Return ONLY valid JSON.

OUTPUT FORMAT: { "seeds": [ {"x": float, "y": float} ] }

INPUT_JSON: { "aspect_ratio": 0.5, "k": 4, "hotspots": [ {"x": 0.85, "y": 0.74, "weight": 0.32, "radius": 0.04}, {"x": 0.98, "y": 0.60, "weight": 0.31, "radius": 0.02}, {"x": 0.32, "y": 0.41, "weight": 0.18, "radius": 0.06} ] }