Good settings
#1
by
YokaiKoibito
- opened
I was getting some rather odd results from this model until I turned Mirostat off and dropped the temperature down to 0.8 or lower — then it was much better behaved. Does anyone else have any tips on good inference settings for this model?