config.ini

[Settings]
temp = 0.23

# Repetitiveness Penalty
# 	Controls how repetitive the AIs output is allowed to be. 
# 	<1 encourages repeats (no one wants this). 
# 	1 is no penalty/off
# 	> 1 penalizes repeats
# 	e.g. 1.2 is a 20% penalty
# 	Common value is 1.2 as it's the default from the CTRL paper who introduced this https://arxiv.org/abs/1909.05858
rep-pen = 1.3

# Repetitiveness Penalty Range
#   Controls how far back tokens are penalized
rep-pen-range = 512

# Repetitiveness Penalty Slope
#   Controls the slope of the penalty curve
rep-pen-slope = 3.33

# The number of words the AI has to choose from.
# 	It always chooses the "top k" most likely next words before randomly picking one according to temperature.
#	  Low values reduce the randomness of the AI similar to temp.
#	  Wont change generation speed. 0 is off
#   Many projects turn this off and use top-p. Original AI Dungeon used 40.
top-keks = 0

# The number of words the AI has to choose from.
#  top-p also called nucleus filtering, keep the top tokens with cumulative probability >= top_p ( see https://arxiv.org/pdf/1904.09751.pdf)
#  similar to top k but probobly better. Can be used together, or you can use this instead.
#  0.9 is used as a default in a wide range of projects and papers
#	 Low values reduce the randomness of the AI similar to temp.
#	 Wont change generation speed
top-p = 0.8

# How long should the longest suggested actions be? higher is slower.
# More technically, this is the number of generated Byte Pair Encoding tokens
# (which are usually whole words) the AI generates for each story response.
generate-num = 40

# Dings the console bell when the AI responds
#	Check your terminal emulator's support for console bells if this doesn't work, it should typically buzz the PC speaker
#	Betcha didn't know ASCII supported sound
console-bell = off

# Maximum width of lines
#	Set to 0 to disable
#	Text wrapping has been much requested since I disabled it from vanilla.
#		In principle this should be a function of your terminal emulator and not an issue
#		Not sure of a good default but 80 was considered an ideal standard number of columns in old PCs.
text-wrap-width = 120

# On means you force use of the cpu even when you have a graphics card. off means you try to use the gpu if you have one
force-cpu = off

# 30 will not spam you with console log message, <30 will spam devs
log-level = 30

# use a dice to decide actions success. E.g. rolling a 1 means "You failed to X"
action-d20 = off

# how many action suggestions to generate, higher is slower
action-sugg = 0

# How weird (and potentially blank and loopy) should the suggested actions be.
# 	0.15 is v conservative, 
# 	0.4 is conservative, 
# 	1.0 is weird (default)
# 	1.5 is glitchy
action-temp = 0.65

# Experimental setting, ignore it for now
top-p-first = on

# Leave "off" unless in Google Colab
colab-mode = off

# Try to enable Python Prompt Toolkit. If problems are detected, it's disabled regardless of the setting
prompt-toolkit = on

# If true, saves after every action, and prompts the user when starting a story what so save it as
autosave = on

# Color scheme that is used if Python Prompt Toolkit is available. A classic-type color scheme can still be used here.
color-scheme = interface/colors-full.ini

# Backup color scheme in case Python Prompt Toolkit isn't available
backup-color-scheme = interface/colors-classic.ini

# Use experimental gpt2 (may be slightly faster, but buggy)
gpt2-experimental = off

# Max number of tokens for GPT-2 Models to use, more = more VRAM but more coherent story
#   Do not set it higher than 1024
history-gpt-2 = 512

# Max number of tokens for GPT-Neo Models to use, more = more VRAM but more coherent story
#   Some people claim setting this a little lower (~2000) is more stable for 8GB VRAM GPUs
#   Do not set it higher than 2048
history-gpt-neo = 2048