rixty
/
SillyTavern
mirror of https://github.com/SillyTavern/SillyTavern.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity
SillyTavern/public/KoboldAI Settings/Deterministic.settings

24 lines
361 B
Plaintext
Raw Normal View History

Llama-2-KoboldAI-presets Here are two presets I've found very useful for Llama 2-based models: - Deterministic takes away the randomness and is good for testing/comparing models because same input equals same output. - Storywriter-Llama2 is the Storywriter preset adjusted for Llama 2's 4K context size. It also works well against Llama 2's repetition/looping issues.
2023-08-01 21:41:34 +02:00
{
"temp": 0,
Update KoboldAI Deterministic.settings After extensive testing, I've adjusted repetition penalty slightly to be the same as simple-proxy-for-tavern's default preset and ooba's LLaMA-Precise settings preset. This fixed some models talking/acting as User.
2023-08-13 19:25:56 +02:00
"rep_pen": 1.18,
Update Deterministic.settings Go back down to 2048 tokens instead of 4096 to be in line with the other non-Llama 2-specific presets
2023-08-03 11:46:53 +02:00
"rep_pen_range": 2048,
Llama-2-KoboldAI-presets Here are two presets I've found very useful for Llama 2-based models: - Deterministic takes away the randomness and is good for testing/comparing models because same input equals same output. - Storywriter-Llama2 is the Storywriter preset adjusted for Llama 2's 4K context size. It also works well against Llama 2's repetition/looping issues.
2023-08-01 21:41:34 +02:00
"top_p": 0,
"top_a": 0,
"top_k": 1,
"typical": 1,
"tfs": 1,
Update KoboldAI Deterministic.settings After extensive testing, I've adjusted repetition penalty slightly to be the same as simple-proxy-for-tavern's default preset and ooba's LLaMA-Precise settings preset. This fixed some models talking/acting as User.
2023-08-13 19:25:56 +02:00
"rep_pen_slope": 0,
Llama-2-KoboldAI-presets Here are two presets I've found very useful for Llama 2-based models: - Deterministic takes away the randomness and is good for testing/comparing models because same input equals same output. - Storywriter-Llama2 is the Storywriter preset adjusted for Llama 2's 4K context size. It also works well against Llama 2's repetition/looping issues.
2023-08-01 21:41:34 +02:00
"sampler_order": [
6,
0,
1,
3,
4,
2,
5
Assorted fixes and improvements (#1208) * Kobold Presets fixed * Help texts fixed * Scale API for connectAPISlash * Quick Reply checkboxes fixed * New Instruct Mode Presets * More date/time macros * ChatML context template and instruct prompt format * Mistral context template and instruct prompt format * Removed use_default_badwordsids from kobold presets * Renamed ChatML to Mistral-OpenOrca (ChatML) * Renamed Mistral-OpenOrca (removed ChatML) * Removed single_line from kobold presets * Removed obsolete use_stop_sequence setting * Ban EOS Token off by default * Split AI Resp. Conf. in global and preset-specific settings * Resolve conflicts * Fix title * Add translations for new help texts * Fix i18n.json whitespace * Make Mistral-OpenOrca system prompt more generic * Renamed "Mistral-OpenOrca" to "ChatML" again * More (UI) fixes and improvements * Sendas hint fixed --------- Co-authored-by: Cohee <18619528+Cohee1207@users.noreply.github.com>
2023-10-07 18:25:36 +02:00
],
"mirostat": 0,
"mirostat_tau": 5,
"mirostat_eta": 0.1,
"grammar": ""
}