The int8 model saved by run_squad can't import by effective_transformer

Open modkzs opened this issue 2 years ago • 0 comments

Branch/Tag/Commit

main

Docker Image Version

nvcr.io/nvidia/tensorflow:20.06-tf1-py3-gcc9-th1.7

GPU name

A100

CUDA Driver

455.23.05

Reproduced Steps

I try to use model export by run_squad and load it into effective_transformer, I got an error saying 'bert/encoder/layer_0/amaxList:0' not exist. Is there any way to save int8 model with amaxList?

Aug 09 '23 05:08 modkzs