R1-Distill-Qwen-14B-Roblox-Luau

A fine tune of deepseek-ai/DeepSeek-R1-Distill-Qwen-14B using boatbomber/roblox-info-dump and boatbomber/the-luau-stack for Roblox domain knowledge.

Recommended inference settings:

Parameter	Value	Notes
System Prompt	You are an expert Roblox developer and Luau software engineer.	Model was fine tuned with this prompt.
temperature	0.5-0.7	Underlying R1 Distill uses this. I've found best results with 0.55.
top_p	0.95	Underlying R1 Distill uses this.

Quantization done using Unsloth.

Available quants:

Quant	Size	Notes
F16	29.55GB	Retains 100% accuracy. Slow and memory hungry.
Q8_O	15.70GB	High resource use, but generally acceptable. Use when accuracy is crucial.
Q6_K	12.12GB	Uses Q6_K for all tensors. Good for high end GPUs.
Q5_K_M	10.51GB	Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q5_K
Q4_K_M	8.99GB	Recommended. Uses Q6_K for half of the attention.wv and feed_forward.w2 tensors, else Q4_K
Q3_K_M	7.34GB	Uses Q4_K for the attention.wv, attention.wo, and feed_forward.w2 tensors, else Q3_K. Quality is noticeably degraded.