Does nnfusion support the partition of a convolution layer?

[baichen318]

Hi all,

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?

Best regards,
Chen