higher context with alpha_value=2.5
#1
by
mclassHF2023
- opened
I tried this with alpha_value=2.5 (at 5_0 bit) and it seems to work surprisingly well at 16k context! So far my favorite Llama3 model!
I tried this with alpha_value=2.5 (at 5_0 bit) and it seems to work surprisingly well at 16k context! So far my favorite Llama3 model!