can we use INT8 WITH plugin layer in tensorrt 4.0？
Anyone follow up on this one? Also interested to know. The plugin sample shows how to support 16-bit or regular float. Does it imply that INT8 calib cannot coexist with plugins?
can we use INT8 WITH plugin layer in tensorrt 4.0？
Anyone follow up on this one? Also interested to know. The plugin sample shows how to support 16-bit or regular float. Does it imply that INT8 calib cannot coexist with plugins?