I come up with several questions, and thanks for your kind comments in advance.
The questions are shown below,
Does nnfusion support partitioning a convolution layer (operator), e.g., a 2D convolution frequently used in CNN?
If multiple continuous layers are partitioned, e.g., multiple continuous convolution layers, how does nnfusion handle the computation dependency between partitioned blocks across different layers?
How does nnfusion synchronize the computation of partitioned blocks?
Hi all,
I come up with several questions, and thanks for your kind comments in advance. The questions are shown below,
nnfusionsupport partitioning a convolution layer (operator), e.g., a 2D convolution frequently used in CNN?nnfusionhandle the computation dependency between partitioned blocks across different layers?nnfusionsynchronize the computation of partitioned blocks?Best regards, Chen