ValueError: Input 0 of node StatefulPartitionedCall/model/batch_normalization/AssignNewValue was passed float from Func/StatefulPartitionedCall/input/_4:0 incompatible with expected resource.

[matthieu-as]

System information

• windows 10
• python 3.6
• tensorflow 2.3.0

When I try to convert my own trained yolov3-tiny model to tflite with the following command: python convert_tflite.py --weights checkpoints/yolov3-tiny-416 --output yolov3.tflite --quantize_mode int8 --dataset coco_dataset/coco/val207.txt , I always hit the fottlowing error: ValueError: Input 0 of node StatefulPartitionedCall/model/batch_normalization/AssignNewValue was passed float from Func/StatefulPartitionedCall/input/_4:0 incompatible with expected resource.

Did anyone run into the same error or knows how to solve it?

[pramodhrachuri]

I am also getting the same error. Were you able to solve it?

[KulkarniKaustubh]

Any luck with this issue? I am facing the same problem.

[piggychu0w0]

I meet the same problem. Did any one solve it?

[WZZDK]

我也遇到这个问题，后面用tf2.3版本的重新运行了save_model.py再使用新生成的文件运行convert_tflite.py就没报错了。

[JonathanSamelson]

In case it helps, I also got this exact error using Linux with a Tensorflow container (2.3.0rc0). Both with convert_rt.py and convert_tflite.py However, I executed save_model.py on another machine and tried to convert this tensorflow model. Instead, I executed save_model.py in the container and it worked.

[ddinhtuann]

You should try to save_model tf again then convert to tflite. It worked for me. I hope it help you all.