Text Generation
GGUF
English
creative
creative writing
fiction writing
plot generation
sub-plot generation
story generation
scene continue
storytelling
fiction story
science fiction
romance
all genres
story
writing
vivid prosing
vivid writing
fiction
roleplaying
bfloat16
swearing
rp
128k context
horror
llama 3.1
mergekit
Inference Endpoints
conversational
Stop Token Shenanigans
#1
by
Anduin1357
- opened
Somehow, token 128008 <|eom_id|> works way better as the EOT token than token 128009 <|eot_id|> and I have no idea why.
Depending on the model's config / tokenizer it can have multiple stop tokens.
I don't full understand it , but it is often a source of model issues.