The Model Parameters is 6243584000. So in dtype float16 the total memory is 6243584000*2byte = 11.63GB.
But using this tool to calculate, same dtype float16 get the total memory is 5.81GB.
Is this an error? Or is there a problem with my calculation method?
The model repository is <THUDM/chatglm3-6b>
@muellerzr
· Sign up or log in to comment