caiovicentino1/polarquant-skills

Naming notice (2026-04-10). The "PolarQuant" technique used in this model is being rebranded to HLWQ (Hadamard-Lloyd Weight Quantization). The change is only the name; the algorithm and the weights in this repository are unchanged.

The rebrand resolves a name collision with an unrelated, earlier KV cache quantization method also named PolarQuant (Han et al., arXiv:2502.02617, 2025). HLWQ addresses weight quantization with a deterministic Walsh-Hadamard rotation and Lloyd-Max scalar codebook; Han et al.'s PolarQuant addresses KV cache quantization with a random polar rotation. The two methods are technically distinct.

Existing loaders that load this repository by ID continue to work without changes. Future model uploads will use the HLWQ name.

Reference paper for this technique: arXiv:2603.29078 (v2 in preparation; v1 still uses the old name).

🧊 PolarQuant Skills + CLI

14 slash commands for Claude Code + 13 CLI commands for terminal.

CLI Install

pip install git+https://github.com/caiovicentino/polarengine-vllm.git[all]

polarquant info google/gemma-4-31B-it
polarquant chat google/gemma-4-31B-it --vision
polarquant quantize google/gemma-4-31B-it --upload
polarquant serve caiovicentino1/model --port 8000
polarquant monitor --stats

Skills Install (Claude Code)

huggingface-cli download caiovicentino1/polarquant-skills --local-dir polarquant-skills
cp polarquant-skills/*.md ~/.claude/commands/

Public Commands

Command	CLI	Skill	Description
Quantize	polarquant quantize	/polarquant	PQ5+INT4 any model
Chat	polarquant chat	/polarquant-inference	Gradio chat (24GB GPUs)
Serve	polarquant serve	/polarquant-serve	OpenAI API server
Info	polarquant info	—	VRAM estimate + specs
Bench	polarquant bench	/polarquant-bench	PQ vs torchao vs BnB
GGUF	polarquant gguf	/polarquant-gguf	ollama/llama.cpp
Monitor	polarquant monitor	/polarquant-monitor	Track downloads
MLX	polarquant mlx	/polarquant-mlx	Apple Silicon
llama.cpp	polarquant llamacpp	/polarquant-llamacpp	KV cache Q3
vLLM KV	polarquant vllm-kv	/polarquant-vllm-kv	vLLM KV compression
Arena	polarquant arena	/polarquant-arena	MMLU, HumanEval
Fine-tune	polarquant finetune	/polarquant-finetune	QLoRA pipeline

Links

📄 Paper · 💻 GitHub · 🤗 Models