just testing for now, qlora merge, several things different between this and the 7b
training
NousResearch/Llama-2-13b-hf tuned on koishi data (without code subsets) for 1 epoch
then tuned on pippa for 1 epoch
then tuned on ludis/geepeetee4 commit 58aabc8 for 1 epoch
then tuned on limarp (without ponyville, lolicit, all the fallen, and eka's portal subsets) Version 2023-09-08 for 2 epochs
all metharme format
prompting
https://rentry.org/tsukasa13b - reccomended prompts and gen settings
The current model version has been trained on prompts using three different roles, which are denoted by the following tokens: <|system|>
, <|user|>
and <|model|>
.
The <|system|>
prompt can be used to inject out-of-channel information behind the scenes, while the <|user|>
prompt should be used to indicate user input. The <|model|>
token should then be used to indicate that the model should generate a response. These tokens can happen multiple times and be chained up to form a conversation history.