TensorflowLite error: Didn't find op for builtin opcode 'SOFTMAX' version '1'. An older version of this builtin might be supported. Are you using an old TFLite binary with a newer model?

General Discussion
, , ,
#1

My envs:
tensorflow version==2.0.0 with cpu
python version==3.7
os: windows10
When I trained a keras model, I used tensorflow lite converter to transfer the code into .tflite, and then use xxd tool to transfer .tflite file into C++ code, in order to deploy the model on the embedded system. But when I compiled the code, it showed the error:
图片
Here is the part of code during the training procedure:

图片
图片793×319 33.7 KB

Thanks so much!

#3

We see that you are using tf version 2.0.0 is not actively supported, please update to stable version tf v2.5, v2.6 and let us know if you are facing same issue.Thanks!

2 Likes
#4

Thanks for your reply!
I installed tensorflow2.6.0 just now, and repeated the whole process, the error was still existed, the same as shown above, why about it?

1 Like
#6

Please check out the installed TensorFlow version in the embedded system. Then, I would recommend using the same TensorFlow versions in the both embedded system and the model conversion.