Language Modeling on TinyLlama 1.1B
0.1Delta PPL (%)SmoothQuant W8A8
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| SmoothQuant W8A8bpw=8.12026.05 | 0.1 | — | — | |
| QAM-W-5.5bpw=5.52026.05 | 0.4 | — | — | |
| GPTQ W4A16 g128bpw=4.12026.05 | 2.6 | — | — | |
| AutoRound W4A16 g128bpw=4.12026.05 | 3.5 | — | — | |
| AWQ W4A16 g128bpw=4.12026.05 | 3.7 | — | — | |
| QAM-W-4bpw=4.02026.05 | 4.2 | — | — | |
| RTN W4A16 g128bpw=4.12026.05 | 5.6 | — | — | |
| QAM-W-polarbpw=4.02026.05 | 6.6 | — | — | |
| DonorTrainable %=18.0%, Time=8.1s2026.01 | — | 54.15 | — | |
| Full fine-tuneTrainable %=100%, Time=104.8s2026.01 | — | 788.36 | — | |
| IA3Trainable %=0.01%, Time=86.0s2026.01 | — | 2,019.41 | — | |
| LoRATrainable %=0.41%, Time=116.5s2026.01 | — | 460.41 | — | |
| Top-KTrainable %=18.0%, Time=41.3s2026.01 | — | 1,157.83 | — | |
| Zero-shotTrainable %=0%2026.01 | — | 1,493.16 | — |