The maximum assistant length of the data set is confirmed to be 97015, but the maximum sequence length of the Qwen2.5-32B-Instruct model appears to be 32768.When fine tuning, was the data set trained as is without any preprocessing?
· Sign up or log in to comment
