BMInf
BMInf copied to clipboard
[FEATURE] Question: why does CPM1.0 need 2 GPU cards while BMInf only need 1 GPU?
where does it optimized?
I have same question, looking forward the answer.
Sorry for the late reply.
BMInf's Key Technologies:
- model quantization
- offloading
We have a technical report that will be made public recently.
Thank you ! Looking forward the report !
report PDF