Support exporting and importing ONNX models
cpetrov opened this issue · 0 comments
cpetrov commented
This can help leverage hardware optimisations and significantly speed up inference: https://huggingface.co/transformers/serialization.html
It seems this would be only possible with transformers>=4.9.0 (see https://huggingface.co/transformers/serialization.html#configuration-based-approach)), so this issue seems related to #3.