"*.tflite" to "*.tmfile" conversion issue.
kprastey opened this issue · comments
Need to run custom mobilenetV2 based "*.tflite" model on the Khadas-VIM3's NPU.
Tried converting a custom mobilenetV2 based classifier model from ".tflite" to ".tmfile", but faced the following error.
It seems like one of the operations (i.e. DEQUANTIZE) isn't supported.
also tried ".tflite" -> ".onnx" ->"*.tmfile". which gave similar errors.
Looking forward to having technical assistance from your end regarding running custom "*.tflite" models having operations like DEQUANTIZE, QUANTIZE, etc. on the khadas VIM3's NPU.
Sorry, We don't have knowledge about this tool. We can support tflite model directly with tensorflowlite + vx-delegate, you can try this approach.
Hey @sunshinemyson,
- It'll be great if you could guide with some documented steps or provide any reference material to help use "tensorflowlite + vx-delegate", especially on the Khadas VIM-3 platform.
- If that's not possible then any platform-agnostic references would also be appreciated.
https://github.com/VeriSilicon/tflite-vx-delegate Can you to work out by follow the README.md firstly?