Update README.md

README.md

@@ -15,8 +15,10 @@ inference: false
 
 Converted from the XORs weights from PygmalionAI's release https://huggingface.co/PygmalionAI/metharme-7b
 
-Quantized for KoboldAI use using https://github.com/0cc4m/GPTQ-for-LLaMa
-
+  Quantized for KoboldAI use using https://github.com/0cc4m/GPTQ-for-LLaMa
+  
+I created several quantized variations of this model and believe this variation to be "best." 
+<b>Benchmark scores at the bottom of readme.</b>
 
 Metharme 7B is an instruct model based on Meta's LLaMA-7B.
 
@@ -85,11 +87,66 @@ As such, it was **not** fine-tuned to be safe and harmless: the base model _and_
 
 <p><strong><font size="5">Benchmarks</font></strong></p>
 
-<p><strong><font size="4">This Model:</font> <br><font size="4">4 Bit</font></strong></p>
+<p><strong><font size="4">This Model:</font> <br><font size="4">4 Bit --Act-order</font></strong></p>
+<strong>Wikitext2</strong>: 6.281311511993408
+
+<strong>Ptb-New</strong>: 46.79158401489258
+
+<strong>C4-New</strong>: 7.906069755554199
+<hr>
+<p><strong><font size="4">4 Bit --True-Sequential --Groupsize 32</font></strong></p>
+<strong>Wikitext2</strong>: 6.424218654632568
+
+<strong>Ptb-New</strong>: 48.48588943481445
+
+<strong>C4-New</strong>: 8.089512825012207
+<hr>
+<p><strong><font size="4">4 Bit --Groupsize 32</font></strong></p>
+<strong>Wikitext2</strong>: 6.902080535888672
+
+<strong>Ptb-New</strong>: 50.237754821777344
+
+<strong>C4-New</strong>: 8.081602096557617
+<hr>
+<p><strong><font size="4">4 Bit --Groupsize 128</font></strong></p>
+<strong>Wikitext2</strong>: 10.206376075744629
+
+<strong>Ptb-New</strong>: 49.00401306152344
 
+<strong>C4-New</strong>: 8.198845863342285
+<hr>
+<p><strong><font size="4">4 Bit --True-Sequential --Groupsize 128</font></strong></p>
+<strong>Wikitext2</strong>: 9.475017547607422
+
+<strong>Ptb-New</strong>: 52.358829498291016
+
+<strong>C4-New</strong>: 8.340700149536133
+<hr>
+<p><strong><font size="4">4 Bit</font></strong></p>
 <strong>Wikitext2</strong>: 6.997134685516357
 
 <strong>Ptb-New</strong>: 58.87525177001953
 
 <strong>C4-New</strong>: 8.824191093444824
-<hr>
+<hr>
+<p><strong><font size="4">4 Bit --True Sequential</font></strong></p>
+<strong>Wikitext2</strong>: 7.735969543457031
+
+<strong>Ptb-New</strong>: 62.92238235473633
+
+<strong>C4-New</strong>: 8.913650512695312
+<hr>
+<p><strong><font size="4">4 Bit --act-order --groupsize 32</font></strong></p>
+<strong>Wikitext2</strong>: 275425.5
+
+<strong>Ptb-New</strong>: 267733.25
+
+<strong>C4-New</strong>: 254506.71875
+<hr>
+<p><strong><font size="4">4 Bit --act-order --groupsize 32</font></strong></p>
+<strong>Wikitext2</strong>: 251321.265625
+
+<strong>Ptb-New</strong>: 250117.859375
+
+<strong>C4-New</strong>: 232929.234375
+<hr>