metadata
license: apache-2.0
base_model: baidu/ERNIE-4.5-300B-A47B-Base-PT
base_model_relation: quantized
quantized_by: turboderp
tags:
- exl3
EXL3 quants of ERNIE-4.5-300B-A47B-Base-PT
2.00 bits per weight
2.10 bits per weight (optimized)
2.25 bits per weight (optimized)
2.50 bits per weight (optimized)
3.00 bits per weight
| Quant | Weights/VRAM | Perplexity | KL-div |
|---|---|---|---|
| 2.00 bpw | 70.2 GB | 4.3711 | 1.1744 |
| 2.10 bpw | 73.4 GB | 1.9047 | 0.4070 |
| 2.25 bpw | 78.6 GB | 1.6274 | 0.2613 |
| 2.50 bpw | 87.8 GB | 1.4719 | 0.1651 |
| 3.00 bpw | 104.9 GB | 1.4358 | 0.1064 |
| Original | 597.1 GB | 1.3199 |