why is onnx_cpu pipeline slower than pytorch_cpu pipeline？

patil-suraj / onnx_transformers

Accelerated NLP pipelines for fast inference on CPU. Built with Transformers and ONNX runtime.

Apache License 2.0

125 stars 27 forks source link

Closed GaoQ1 closed 3 years ago

GaoQ1 commented 3 years ago

I run the ipynb on my computer but onnx_cpu pipeline is slower than pytorch_cpu pipeline