parameters guide
samplers guide
model generation
role play settings
quant selection
arm quants
iq quants vs q quants
optimal model setting
gibberish fixes
coherence
instructing following
quality generation
chat settings
quality settings
llamacpp server
llamacpp
lmstudio
sillytavern
koboldcpp
backyard
ollama
model generation steering
steering
model generation fixes
text generation webui
ggufs
exl2
full precision
quants
imatrix
neo imatrix
Update README.md
Browse files
README.md
CHANGED
@@ -48,7 +48,7 @@ Even if you are not using my models, you may find this document useful for any m
|
|
48 |
|
49 |
If you are currently using model(s) that are difficult to "wrangle" then apply "Class 3" or "Class 4" settings to them.
|
50 |
|
51 |
-
This document will be updated over time too.
|
52 |
|
53 |
Please use the "community tab" for suggestions / edits / improvements.
|
54 |
|
@@ -614,7 +614,7 @@ You may therefore want to experiment to with dropping the settings (SLOWLY) for
|
|
614 |
|
615 |
<B>DRY:</B>
|
616 |
|
617 |
-
Dry affects repetition (and repeat "penalty") at the word, phrase, sentence and even paragraph level. Read about "DRY" above, in the "Additional Links" links section above.
|
618 |
|
619 |
Class 3:
|
620 |
|
@@ -630,12 +630,16 @@ dry_multiplier: .8 to 1.12+
|
|
630 |
|
631 |
dry_allowed_length: 2 (or less)
|
632 |
|
633 |
-
dry_base: 1.15 to 1.
|
|
|
|
|
|
|
|
|
634 |
|
635 |
|
636 |
<B>QUADRATIC SAMPLING:</B>
|
637 |
|
638 |
-
This sampler alters the "score" of ALL TOKENS at the time of generation. See "Additional Links" links section above for more information.
|
639 |
|
640 |
Class 3:
|
641 |
|
@@ -649,6 +653,11 @@ smoothing_factor: 3 to 5 (or higher)
|
|
649 |
|
650 |
smoothing_curve: 1.5 to 2.
|
651 |
|
|
|
|
|
|
|
|
|
|
|
652 |
|
653 |
<B>ANTI-SLOP - Kolbaldcpp only</B>
|
654 |
|
|
|
48 |
|
49 |
If you are currently using model(s) that are difficult to "wrangle" then apply "Class 3" or "Class 4" settings to them.
|
50 |
|
51 |
+
This document will be updated over time too and is subject to change without notice.
|
52 |
|
53 |
Please use the "community tab" for suggestions / edits / improvements.
|
54 |
|
|
|
614 |
|
615 |
<B>DRY:</B>
|
616 |
|
617 |
+
Dry ("Don't Repeat Yourself") affects repetition (and repeat "penalty") at the word, phrase, sentence and even paragraph level. Read about "DRY" above, in the "Additional Links" links section above.
|
618 |
|
619 |
Class 3:
|
620 |
|
|
|
630 |
|
631 |
dry_allowed_length: 2 (or less)
|
632 |
|
633 |
+
dry_base: 1.15 to 1.75+
|
634 |
+
|
635 |
+
Dial the "dry_muliplier" up or down to "reign in" or "release the madness" so to speak from the core model.
|
636 |
+
|
637 |
+
For Class 4 models this is used to control some of the model's bad habit(s).
|
638 |
|
639 |
|
640 |
<B>QUADRATIC SAMPLING:</B>
|
641 |
|
642 |
+
This sampler alters the "score" of ALL TOKENS at the time of generation and as a result affects the entire generation of the model. See "Additional Links" links section above for more information.
|
643 |
|
644 |
Class 3:
|
645 |
|
|
|
653 |
|
654 |
smoothing_curve: 1.5 to 2.
|
655 |
|
656 |
+
Dial the "smoothing factor" up or down to "reign in" or "release the madness" so to speak.
|
657 |
+
|
658 |
+
In Class 3 models, this has the effect of modifying the prose closer to "normal" with as much or little (or a lot!) touch of "madness" from the root model.
|
659 |
+
|
660 |
+
In Class 4 models, this has the effect of modifying the prose closer to "normal" with as much or little (or a lot!) touch of "madness" from the root model AND wrangling in some of the core model's bad habits.
|
661 |
|
662 |
<B>ANTI-SLOP - Kolbaldcpp only</B>
|
663 |
|