Thai Nguyen
Thai Nguyen
@Xhark Could you take a look.
@ebrevdo I think the BroadcastArgs may come from using broadcast_to on a dynamic tensor. I am working on supporting BroadcastArgs now.
Unfortunately, it is still under review.
@soldierofhell BroadcastArgs is added to master branch. You could try it using the nightly now.
error: illegal scale: INF That probably mean there is INF value in one tensor in the model.
It seems not straight forward now, the model is converted to tflite before quantization so we might have to add log somewhere in the calibration part. I wonder, there seems...
I see. @sngyhan TTS model will require dynamic-range quantization, right? Can you guide @zhuyu14 how to do that?
Generally look good to me. Only a nit that I think the comments in transpose_conv.cc and conv.cc should be kept (just move to the right place). After that I'll happy...
Can you add a test in tensorflow/lite/testing/op_tests/tile.py with "fully_quantize" equals true.
I am OK with the kernel changes but would like to have someone's opinion on tensorflow/lite/tools/optimize/operator_property.cc.