Qwen2.5-14B-Instruct-GGUF / perplexity.md
ThomasBaruzier's picture
Upload perplexity.md
78b6e56 verified
|
raw
history blame
794 Bytes

Qwen2.5-14B-Instruct Quant Size (MB) PPL Size (%) Accuracy (%) PPL error rate IQ1_S 3441 22.0082 0.16818 IQ1_M 3693 15.0790 0.11060 IQ2_XXS 4114 9.6047 0.06625 IQ2_XS 4487 8.3649 0.05574 IQ2_S 4772 8.1942 0.05480 IQ2_M 5109 7.7261 0.05177 Q2_K_S 5148 8.0641 0.05490 Q2_K 5504 7.6005 0.05146 IQ3_XXS 5672 6.9285 0.04547 IQ3_XS 6088 6.7210 0.04329 Q3_K_S 6352 6.8697 0.04576 IQ3_S 6383 6.6246 0.04285 IQ3_M 6597 6.6359 0.04256 Q3_K_M 7000 6.5281 0.04300 Q3_K_L 7558 6.4323 0.04211 IQ4_XS 7744 6.2005 0.04022 Q4_0 8149 6.2928 0.04095 IQ4_NL 8154 6.2080 0.04032 Q4_K_S 8177 6.1630 0.03976 Q4_K_M 8572 6.1311 0.03957 Q4_1 8958 6.1674 0.03981 Q5_K_S 9791 6.0411 0.03886 Q5_0 9817 6.0504 0.03895 Q5_K_M 10023 6.0389 0.03888 Q5_1 10625 6.0366 0.03885