Optimized Vector Matrix Product #872
Description
We should create a kernel that is as fast as cuBLAS on all shapes. Really important for LLM inference.
{{ message }}
We should create a kernel that is as fast as cuBLAS on all shapes. Really important for LLM inference.