Text Generation
Transformers
Safetensors
English
llama
llama-factory
Not-For-All-Audiences
conversational
text-generation-inference
Inference Endpoints
aaronday3 commited on
Commit
cd20a2a
1 Parent(s): ee82e2b

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +4 -0
README.md CHANGED
@@ -93,6 +93,10 @@ If you want SFW, remove all system messages including provided one. In this mode
93
  As said, if instruct refusal prefill 2-3 words. **Otherwise we deliberately trained the model to sometimes refuse romantic advances, this is more realistic.** <br>
94
  If you don't like it, **you can override** by editing the character message and continue RP.
95
  <br>
 
 
 
 
96
  <h2>Other Important Tips</h2>
97
 
98
  Take active role in the RP and say the type of response you expect. You don't always have to do this, but it helps sometimes. For example instead of *we drink and drink 15 glasses of champagne* say *we drink and drink 15 glasses of champagne, both becoming extremely drunk*
 
93
  As said, if instruct refusal prefill 2-3 words. **Otherwise we deliberately trained the model to sometimes refuse romantic advances, this is more realistic.** <br>
94
  If you don't like it, **you can override** by editing the character message and continue RP.
95
  <br>
96
+ <h2>RoPE - 16K Context</h2>
97
+
98
+ I have tested these settings and they work OK for 16K. Depending on the roleplay complexity and message length, experiment with if the model starts breaking or not. For me, 16K works fine.
99
+ <img src="https://cdn-uploads.huggingface.co/production/uploads/630cf5d14ca0a22768bbe10c/3f7JOEnXhKCDcDF4Eiq-B.png" alt="" width="300"/>
100
  <h2>Other Important Tips</h2>
101
 
102
  Take active role in the RP and say the type of response you expect. You don't always have to do this, but it helps sometimes. For example instead of *we drink and drink 15 glasses of champagne* say *we drink and drink 15 glasses of champagne, both becoming extremely drunk*