LLM Quantization on Llama-2-70B
2.2GPU Hours (h)SignRoundV1
Evaluation results
| Method | Links | |
|---|---|---|
| SignRoundV1One A100-80GB?=true2025.12 | 2.2 | |
| SignRoundV2One A100-80GB?=true2025.12 | 2.5 | |
| SignRoundV2 (enhanced)One A100-80GB?=true, enhanced_version=true2025.12 | 6 | |
| EfficientQATOne A100-80GB?=true2024.07 | 41 | |
| EfficientQATOne A100-80GB?=true2025.12 | 41 | |
| BitDistillerOne A100-80GB?=false2024.07 | 64 | |
| DB-LLMOne A100-80GB?=false2024.07 | 82 | |
| PB-LLMOne A100-80GB?=false2024.07 | 90 | |
| QuIP#One A100-80GB?=false2025.12 | 270 | |
| QuiP#One A100-80GB?=false2024.07 | 300 | |
| AQLMOne A100-80GB?=true2024.07 | 336 | |
| AQLMOne A100-80GB?=true2025.12 | 336 | |
| LLM-QATOne A100-80GB?=false2024.07 | 900 |