[Bugfix][Relay][Keras] Fix the wrong implementation logic about cropping2D

[jikechao]

This PR fixed the wrong calculation logic which will lead to wrong inference results.
The original cropping algorithm shown below is incorrect.

I modified the code to fix it. Meanwhile, a test case that can trigger this bug was added to the unit test.

Step to reproduce

import tvm
import tvm.relay as relay
import numpy as np
from tensorflow import keras
from tensorflow.keras import layers, models
input_shape = (2, 9, 9, 2)
input_data = np.random.random(size=input_shape)
x = layers.Input(shape=input_shape[1:], dtype='float32')

layer = keras.layers.Cropping2D(cropping=[2,1], data_format='channels_last')
layer.set_weights(layer.get_weights())

y = layer(x)
model = models.Model(x, y)
model.summary()
res_keras = model(input_data)

shape_dict = {'input_1': input_shape}
mod, params = relay.frontend.from_keras(model, shape_dict)
with tvm.transform.PassContext(opt_level=3):
    model = relay.build_module.create_executor("graph", mod, tvm.cpu(0), 'llvm', params).evaluate()

test_x_tvm = input_data
res_tvm = model(tvm.nd.array(test_x_tvm.astype('float32'))).numpy()

np.testing.assert_allclose(res_keras, res_tvm, atol=1e-3, rtol=1e-3)

cc @echuraev @masahi @Hzfengsy

[tvm-bot]

Thanks for contributing to TVM! Please refer to the contributing guidelines https://tvm.apache.org/docs/contribute/ for useful information and tips. Please request code reviews from Reviewers by @-ing them in a comment.

• cc @shingjan _{See #10317 for details}

_{Generated by tvm-bot}

[masahi]

See #14983 (comment) for the arm build issue.

[masahi]

@tvm-bot rerun

[jikechao]

The initial patch will lead to existing test cases crashing. So, I updated the patch to fix it.

[jikechao]

@tvm-bot rerun

[jikechao]

@tvm-bot rerun

[jikechao]

@echuraev @masahi CI test has finished. Could you help me to merge this PR? Thanks!