Are there plans to support LSTM or GRU on the edge TPU?
@ankandrew Thanks for reaching out!
Yes, we are planning on this and started experimenting with it.
FYI, tflite conversion for LSTM model is quite new for the tensorflow team and currently quantization of the model is still not available.
You can check here for the tflite roadmap: https://www.tensorflow.org/lite/guide/roadmap
@Namburger I'm wondering if there are any updates on this given that some time has passed.
It seems that tflite still doesn't support full integer quantization for LSTMs but has added float16 support, which I realize doesn't help for running on an edge tpu... but maybe some progress has still been made?
Hello, I have the same problem. Is there any timeline that i can follow? Or just update tensorflow from time to time and try to convert and see if it works?
Most helpful comment
Hello, I have the same problem. Is there any timeline that i can follow? Or just update tensorflow from time to time and try to convert and see if it works?