Finaly, merged some loras and work.

1. wizardLM-unsensored + finnish + Germany + Italian + esJoke. 7B

2. wizardLM-unsensored + StarCoder(lora.) 13B

for serious, I found out it could be worse to merge more loras without test. 

it's better be tested in llama.cpp.





---------------------------
sorry for messed up.

since one pointed out its no different with the wizard model.

i tested it with same seed same ctx same input prompts sequence. 

theres NO DIFFERENCE between that model and the original wizardlm.

i guess i need to review the all process. 
 
sorry again.
------------------------------ 




Be carefully since it's uncensored, it's your responsiblility to take it seriouslly, 
if you download that model, you do agree not using it in any circumstance leads to hurt anything or anybody include yourself in any dimensions ( not only physical but also mental).

thank you!