Datasets:

Tags:
Not-For-All-Audiences
roleplay
conversational
Libraries:
Datasets
License:
IkariDev commited on
Commit
e0d02df
1 Parent(s): 416221c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -1
README.md CHANGED
@@ -10,8 +10,13 @@ size_categories:
10
 
11
  ## MinervaAI is proud to present its very first public dataset release: Aesir-Preview
12
 
 
 
13
  Contained within this ShareGPT-based dataset are 1000 fully synthetic roleplay dialogue generations between an anonymous user and character cards from Chub.ai, the latter which were carefully checked, corrected and improved upon.
14
 
15
  Each generation is the results of dozens of automated validations, corrections and manual curations to ensure they're of the highest quality that can be achieved within the limitations of the model used, which was GPT 3.5 Instruct.
16
 
17
- ⚠️ **WARNING:** This dataset is filled to the brim with NSFW data and contains a wide variety of erotic themes, potentially disturbing scenes and very strong language. Models trained on this data will therefore be heavily biased towards recreating such behaviour.
 
 
 
 
10
 
11
  ## MinervaAI is proud to present its very first public dataset release: Aesir-Preview
12
 
13
+ ⚠️ **WARNING:** This is a preview dataset and may not reflect the content or quality of the final result. Use discretion and caution when accessing or utilizing this data.
14
+
15
  Contained within this ShareGPT-based dataset are 1000 fully synthetic roleplay dialogue generations between an anonymous user and character cards from Chub.ai, the latter which were carefully checked, corrected and improved upon.
16
 
17
  Each generation is the results of dozens of automated validations, corrections and manual curations to ensure they're of the highest quality that can be achieved within the limitations of the model used, which was GPT 3.5 Instruct.
18
 
19
+ ⚠️ **NSFW WARNING:** This dataset is filled to the brim with NSFW data and contains a wide variety of erotic themes, potentially disturbing scenes and very strong language. Models trained on this data will therefore be heavily biased towards recreating such behaviour.
20
+
21
+
22
+ By Gryphe, Doctor Shotgun, IkariDev, Undi, Mixel, [Chat Error], kubernetes_bad and StefanGliga