Language Modeling on WikiText-2 (CTX=2048)
6.1PerplexityLiftQuant-24/10
Evaluation Results
| Method | Links | |
|---|---|---|
| LiftQuant-24/10Model=Llama-2-7B, BITS=2.40, Search time (S.T.)=1S2026.06 | 6.1 | |
| QTIPModel=Llama-2-7B, Coding=TCQ, BITS=2.002026.06 | 6.28 | |
| LiftQuant-30/14Model=Llama-2-7B, BITS=2.14, Search time (S.T.)=4S2026.06 | 6.3 | |
| LiftQuant-32/16Model=Llama-2-7B, BITS=2.00, Search time (S.T.)=4S2026.06 | 6.53 | |
| LiftQuant-16/8Model=Llama-2-7B, BITS=2.00, Search time (S.T.)=<<0.1S2026.06 | 6.6 | |
| QUIP#Model=Llama-2-7B, Coding=E8, BITS=2.002026.06 | 6.6 | |
| INT2Model=Llama-2-7B, BITS=2.002026.06 | 7.62 | |
| LiftQuant-32/20Model=Llama-2-7B, BITS=1.60, Search time (S.T.)=0.3S2026.06 | 7.71 |