any contamination results?
#4
by
wukongai
- opened
in my test, the test results were very unsatisfactory.
btw: i use the q8 gguf.
I have conducted a number of tests, some more objective than others, and have found this model to outperform Mixtral 8x7B (AWQ as well as via DeepInfra), NeuralBeagle14-7B and Pluto 24B DPO 200. However, I am running this in full-precision. My tests involve sentiment analysis, classification and text extraction.
Nevertheless, I too would like to see contamination results for this model ( @yunconglong ) and its parent ( @TomGrc ).
I just trained DPO with jondurbin/truthy-dpo-v0.1, it is very samll dataset.I'm sure no contamination results during DPO training.