gptq
Post-training 4-bit quantization for LLMs with minimal accuracy loss. Use for deploying large models (70B, 405B) on consumer GPUs, when you need 4× memory reduction with <2% perplexity degradation, or for faster inference (3-4× speedup) vs FP16. Integrates with transformers and PEFT for QLoRA fine-tuning.
| Field | Value |
|---|---|
| Owner | Orchestra-Research |
| Curator | Thomas Cherickal |
| Source | Orchestra-Research/AI-Research-SKILLs |
| Repo stars | 14,542 |
| License | MIT |
| Repo last push | 2026-10-11 |
Links
- Roles: local-ai-engineer
- Source repo: Orchestra-Research/AI-Research-SKILLs
- Same repo: axolotl, deepspeed, distributed-llm-pretraining-torchtitan, fine-tuning-openvla-oft, fine-tuning-serving-openpi, awq-quantization, gguf-quantization, hqq-quantization
- Registry seed skills for this role: local-ai-engineer seed skills