meta-llama

/

Llama-3.1-8B-Instruct

like 3.37k

Follow

Meta Llama 17.2k

Hi everyone,
I am using Llama 3.1: 8b-instruct.
I am trying to get the exact same response, so I set these parameters with the following values:

"temperature": 0,  
"top_p": 1,  
"top_k": 1,  
"frequency_penalty": 0,  
"presence_penalty": 0,  
"repetition_penalty": 10,  
"min_p": -1,  
"top_a": 0,  

However, it still gives me different responses (it less but still there a difference). Any ideas?
Thanks!