Jae sung Chung comments

Results 28 comments of


                                            Jae sung Chung

[TFLite] Quantization and unfolding of the ADD_N operator

This is a good example of quantization aware conversion. FYI, @karimnosseir @renjie-liu

[TFLite] Quantization and unfolding of the ADD_N operator

I think that is the expected behavior. The TF Lower pass will be trigged only when the given TensorFlow core operators are not directly legalized. Since we have the ADD_N...

[TFLite] Quantization and unfolding of the ADD_N operator

For the workaround, maybe you create a custom keras layer to meet your needs.

[TFLite] Quantization and unfolding of the ADD_N operator

@teijeong could you triage this issue?

[TFLite] Quantization and unfolding of the ADD_N operator

FYI, @jianlijianli

[TFLite] Quantization and unfolding of the ADD_N operator

TFLite converter is a gateway to the TFLite product. However, it contains several components, TF -> TFLite graph lowering and model optimization techniques. This issue is related to the model...

TFLite converter does not convert dilated conv to single conv op when spatial dimension is dynamic

For dynamic dimensions, we can't make sure that the given inputs always are supported with the dilated conversion lowering. We have seen some cases that are not correctly performed with...

TFLite converter does not convert dilated conv to single conv op when spatial dimension is dynamic

Here are some test cases related to that. https://github.com/tensorflow/tensorflow/blob/a30d20b632b4ffbfd437ccf8ee205fef0917a3eb/tensorflow/compiler/mlir/lite/tests/dilated-conv.mlir#L329