[TFLite] Fix _test_tflite2_quantized_depthwise_convolution is unused

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.

[blackkker]

1. As [Bug] _test_tflite2_quantized_depthwise_convolution is unused #11356 said. The entry of the _test_tflite2_quantized_depthwise_convolution has been completed.

2. And an input_node error has also been fixed, because the input name of the tflite model quantized by the _quantize_keras_model() will be rewritten(no longer data_in.name.replace(":0", ""))

[blackkker]

cc @leandron, please have a look, thanks!

[lhutton1]

also cc @NicolaLancellotti(as this might conflict with #12130)

[blackkker]

Also, these lines in test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 4590 to 4593 in eb7cf70

	ifpackage_version.parse(tf.VERSION) <package_version.parse("1.15.0"):
	pytest.skip("Unsupported in tflite < 1.15.0")
	else:
	pytest.skip("This segfaults with tensorflow 1.15.2 and above")

disable the following tests below test_forward_qnn_mobilenet_v3_net()

tvm/tests/python/frontend/tflite/test_forward.py

Lines 5106 to 5113 in eb7cf70

	test_forward_qnn_coco_ssd_mobilenet_v1()
	# TFLite 2.1.0 quantized tests
	test_forward_tflite2_qnn_resnet50()
	test_forward_tflite2_qnn_inception_v1()
	test_forward_tflite2_qnn_mobilenet_v2()
	test_forward_tflite_float16()

Can i put test_forward_qnn_mobilenet_v3_net() to the bottom？

[blackkker]

also cc @NicolaLancellotti(as this might conflict with #12130)

I think @NicolaLancellotti 's #12130 has fixed the input name error, great!

[blackkker]

@AndrewZhaoLuo

[AndrewZhaoLuo]

I will take a look tomorrow

[blackkker]

I will remove fix input_node name error when #12130 merged.
Beacuse https://github.com/NicolaLancellotti/tvm/blob/6a042e0a9e5d5290a51a4cf308dd144535d441ad/tests/python/frontend/tflite/test_forward.py#L938-L941
has fixed the input name error with easier way.

[AndrewZhaoLuo]

So what I understand is the original tests in the issue were disabled since they were flaky. This issue seems closed now #6064 so it's fine to re-enable tests. However, due to some name change some additional code needs to be added to make the tests pass?

Seems fine to me, thanks for contribution.