Naphula/Updated_Settings
Recommended Settings In particular temp 1 and topnsigma 1.25 seems to help a lot with improving quality but sometimes you need to reswipe. Some models are more sensitive to Rep Pen settings, 1.12 is another, you may need to adjust higher for smaller models (like 1.4 if overcooked) or lower for smarter ones. Not sure about adaptive_p that is a newer setting added to kobold Ban em dashes and ellipses (or set to -50 logit bias) to reduce slop: —||$||...||$||… (bolded kobold… See the full description on the dataset page: https://huggingface.co/datasets/Naphula/Updated_Settings.
470
1*.7z filter=lfs diff=lfs merge=lfs -text2*.arrow filter=lfs diff=lfs merge=lfs -text3*.avro filter=lfs diff=lfs merge=lfs -text4*.bin filter=lfs diff=lfs merge=lfs -text5*.bz2 filter=lfs diff=lfs merge=lfs -text6*.ckpt filter=lfs diff=lfs merge=lfs -text7*.ftz filter=lfs diff=lfs merge=lfs -text8*.gz filter=lfs diff=lfs merge=lfs -text9*.h5 filter=lfs diff=lfs merge=lfs -text10*.joblib filter=lfs diff=lfs merge=lfs -text11*.lfs.* filter=lfs diff=lfs merge=lfs -text12*.lz4 filter=lfs diff=lfs merge=lfs -text13*.mds filter=lfs diff=lfs merge=lfs -text14*.mlmodel filter=lfs diff=lfs merge=lfs -text15*.model filter=lfs diff=lfs merge=lfs -text16*.msgpack filter=lfs diff=lfs merge=lfs -text17*.npy filter=lfs diff=lfs merge=lfs -text18*.npz filter=lfs diff=lfs merge=lfs -text19*.onnx filter=lfs diff=lfs merge=lfs -text20*.ot filter=lfs diff=lfs merge=lfs -text21*.parquet filter=lfs diff=lfs merge=lfs -text22*.pb filter=lfs diff=lfs merge=lfs -text23*.pickle filter=lfs diff=lfs merge=lfs -text24*.pkl filter=lfs diff=lfs merge=lfs -text25*.pt filter=lfs diff=lfs merge=lfs -text26*.pth filter=lfs diff=lfs merge=lfs -text27*.rar filter=lfs diff=lfs merge=lfs -text28*.safetensors filter=lfs diff=lfs merge=lfs -text29saved_model/**/* filter=lfs diff=lfs merge=lfs -text30*.tar.* filter=lfs diff=lfs merge=lfs -text31*.tar filter=lfs diff=lfs merge=lfs -text32*.tflite filter=lfs diff=lfs merge=lfs -text33*.tgz filter=lfs diff=lfs merge=lfs -text34*.wasm filter=lfs diff=lfs merge=lfs -text35*.xz filter=lfs diff=lfs merge=lfs -text36*.zip filter=lfs diff=lfs merge=lfs -text37*.zst filter=lfs diff=lfs merge=lfs -text38*tfevents* filter=lfs diff=lfs merge=lfs -text39# Audio files - uncompressed40*.pcm filter=lfs diff=lfs merge=lfs -text41*.sam filter=lfs diff=lfs merge=lfs -text42*.raw filter=lfs diff=lfs merge=lfs -text43# Audio files - compressed44*.aac filter=lfs diff=lfs merge=lfs -text45*.flac filter=lfs diff=lfs merge=lfs -text46*.mp3 filter=lfs diff=lfs merge=lfs -text47*.ogg filter=lfs diff=lfs merge=lfs -text48*.wav filter=lfs diff=lfs merge=lfs -text49# Image files - uncompressed50*.bmp filter=lfs diff=lfs merge=lfs -text51*.gif filter=lfs diff=lfs merge=lfs -text52*.png filter=lfs diff=lfs merge=lfs -text53*.tiff filter=lfs diff=lfs merge=lfs -text54# Image files - compressed55*.jpg filter=lfs diff=lfs merge=lfs -text56*.jpeg filter=lfs diff=lfs merge=lfs -text57*.webp filter=lfs diff=lfs merge=lfs -text58# Video files - compressed59*.mp4 filter=lfs diff=lfs merge=lfs -text60*.webm filter=lfs diff=lfs merge=lfs -text61