Create README.md

README.md

@@ -0,0 +1,14 @@
+---
+license: mit
+language:
+- en
+---
+
+My own quantizations.  
+output and embed tesnors quantized to f16.  
+all other tensors quantized to q5_k or q6_k.  
+the q8_0 version is pure (all tensors quantized to Q8_0 just for reference)  
+
+Result:  
+both f16.q6 and f16.q5 are smaller than q8_0 standard quantization  
+and they perform as well as the pure f16.