Update Dynamic-Temp sampler
Browse files- README.md +17 -2
- Samplers/[Dynamic-Temp]Roleplay.json +3 -3
README.md
CHANGED
@@ -46,7 +46,7 @@ Or this:
|
|
46 |
|
47 |
**I have decided to remove old samplers and only keep basic presets, I want people to play around and find what works best for them. Change context to desired context length**
|
48 |
|
49 |
-
**[Dynamic-Temp]Roleplay
|
50 |
|
51 |
[SillyTavern Docs](https://docs.sillytavern.app/usage/common-settings/#sampler-parameters)
|
52 |
|
@@ -55,12 +55,17 @@ Feel free to play with this one, lower values are more grounded.
|
|
55 |
|
56 |
**Min-P**
|
57 |
Higher values chop off more probabilities.
|
|
|
|
|
58 |
Values between 0.025 - 0.10 are good, personally I would use 0.075 or lower.
|
59 |
|
60 |
**Repetition Penalty**
|
61 |
Tries to decrease repetition.
|
|
|
|
|
62 |
Do not set it higher than 1.2.
|
63 |
|
|
|
64 |
1.05 - 1.15 seem to work fine.
|
65 |
|
66 |
**Rep Pen Range**
|
@@ -74,12 +79,22 @@ Decreases repetition.
|
|
74 |
Increases word variety.
|
75 |
|
76 |
**Dynamic Temperature**
|
77 |
-
I have it disables at the moment.
|
78 |
Min and Max temps, free to change as desired.
|
|
|
|
|
79 |
Exponent, do not set Exponent higher than the default of 1.
|
|
|
|
|
80 |
You might want to try playing around and setting it lower than 1, this pushes lower probabilies higher.
|
81 |
|
|
|
|
|
|
|
82 |
**Smooth Sampling**
|
83 |
This one is great, smoothens out probabilities.
|
|
|
|
|
84 |
Lower is more diverse.
|
|
|
|
|
85 |
Recommended range 0.1 - 0.3
|
|
|
46 |
|
47 |
**I have decided to remove old samplers and only keep basic presets, I want people to play around and find what works best for them. Change context to desired context length**
|
48 |
|
49 |
+
**[Dynamic-Temp]Roleplay has a tendency to break asterisk formatting, however is also stops alot of the GPTism.**
|
50 |
|
51 |
[SillyTavern Docs](https://docs.sillytavern.app/usage/common-settings/#sampler-parameters)
|
52 |
|
|
|
55 |
|
56 |
**Min-P**
|
57 |
Higher values chop off more probabilities.
|
58 |
+
|
59 |
+
|
60 |
Values between 0.025 - 0.10 are good, personally I would use 0.075 or lower.
|
61 |
|
62 |
**Repetition Penalty**
|
63 |
Tries to decrease repetition.
|
64 |
+
|
65 |
+
|
66 |
Do not set it higher than 1.2.
|
67 |
|
68 |
+
|
69 |
1.05 - 1.15 seem to work fine.
|
70 |
|
71 |
**Rep Pen Range**
|
|
|
79 |
Increases word variety.
|
80 |
|
81 |
**Dynamic Temperature**
|
|
|
82 |
Min and Max temps, free to change as desired.
|
83 |
+
|
84 |
+
|
85 |
Exponent, do not set Exponent higher than the default of 1.
|
86 |
+
|
87 |
+
|
88 |
You might want to try playing around and setting it lower than 1, this pushes lower probabilies higher.
|
89 |
|
90 |
+
|
91 |
+
When setting exponent lower than 1, set Min-P a little higher (0.075)
|
92 |
+
|
93 |
**Smooth Sampling**
|
94 |
This one is great, smoothens out probabilities.
|
95 |
+
|
96 |
+
|
97 |
Lower is more diverse.
|
98 |
+
|
99 |
+
|
100 |
Recommended range 0.1 - 0.3
|
Samplers/[Dynamic-Temp]Roleplay.json
CHANGED
@@ -8,7 +8,7 @@
|
|
8 |
"epsilon_cutoff": 0,
|
9 |
"eta_cutoff": 0,
|
10 |
"typical_p": 1,
|
11 |
-
"min_p": 0.
|
12 |
"rep_pen": 1.05,
|
13 |
"rep_pen_range": 2048,
|
14 |
"no_repeat_ngram_size": 0,
|
@@ -23,8 +23,8 @@
|
|
23 |
"early_stopping": false,
|
24 |
"dynatemp": true,
|
25 |
"min_temp": 1,
|
26 |
-
"max_temp": 1.
|
27 |
-
"dynatemp_exponent": 0.
|
28 |
"smoothing_factor": 0,
|
29 |
"smoothing_curve": 1,
|
30 |
"add_bos_token": true,
|
|
|
8 |
"epsilon_cutoff": 0,
|
9 |
"eta_cutoff": 0,
|
10 |
"typical_p": 1,
|
11 |
+
"min_p": 0.075,
|
12 |
"rep_pen": 1.05,
|
13 |
"rep_pen_range": 2048,
|
14 |
"no_repeat_ngram_size": 0,
|
|
|
23 |
"early_stopping": false,
|
24 |
"dynatemp": true,
|
25 |
"min_temp": 1,
|
26 |
+
"max_temp": 1.15,
|
27 |
+
"dynatemp_exponent": 0.33,
|
28 |
"smoothing_factor": 0,
|
29 |
"smoothing_curve": 1,
|
30 |
"add_bos_token": true,
|