Closed ohjunee closed 1 year ago
why did the onnx optimizer name the factor for optimization "pass"? Is there a reason ??
It is from the compiler field. Initially a "pass" means traversing the computation graph once, now people use it to describe something like pattern matching.
I want to do an impact measurement by layer/operator (fused layer/module) to see how it affects the layer optimized with onnx optimizer, is there any evidence that it improves performance?
you might want to use onnxsim, which is a more powerful tool based on onnx optimizer
closing. feel free to reopen it if you have any further questions
why did the onnx optimizer name the factor for optimization "pass"? Is there a reason ??
I want to do an impact measurement by layer/operator (fused layer/module) to see how it affects the layer optimized with onnx optimizer, is there any evidence that it improves performance?