Quantization Distribution Evaluation on Wiki2 (calibration set)
0KL DivergenceBF16 (baseline)
Evaluation Results
| Method | Links | |
|---|---|---|
| BF16 (baseline)Precision=BF162026.05 | 0 | |
| PO2(Split87)Quantization=W4A42026.05 | 0.0813 | |
| PO2(NF4)Quantization=W4A42026.05 | 0.0893 | |
| Split87Quantization=W4A42026.05 | 0.0897 | |
| BOF4Quantization=W4A42026.05 | 0.0969 | |
| NF4Quantization=W4A42026.05 | 0.0988 | |
| IF4Quantization=W4A42026.05 | 0.0992 | |
| SFP4Quantization=W4A42026.05 | 0.1019 | |
| 4over6Quantization=W4A42026.05 | 0.1023 | |
| NVFP4Quantization=W4A42026.05 | 0.1143 | |
| NVINT4Quantization=W4A42026.05 | 0.1429 |