c94623ad0dac5fa5ce7a9a3c4ecb794351ebc610,tests/python/frontend/tflite/test_forward.py,,_test_relu,#Any#Any#,2870

Before Change

            pytest.skip("Testcase requires tflite version >= 2.1.0")
        data_in = tf.keras.layers.Input(shape=data.shape[1:])
        relu = tf.keras.layers.ReLU()(data_in)
        keras_model = tf.keras.models.Model(inputs=data_in, outputs=relu)
        input_name = data_in.name.split(":")[0]

        // To create quantized values with dynamic range of activations, needs representative dataset
        def representative_data_gen():

After Change

            inq_data = tf.quantization.fake_quant_with_min_max_args(
                in_data, min=-10, max=10, name="inq_0"
            )
            input_range = {"inq_0": (-10, 10)}
            out = nn_ops.relu(inq_data)
            out = tf.quantization.fake_quant_with_min_max_args(out, min=0, max=6, name="out")
            compare_tflite_with_tvm(
                data, "inq_0:0", [inq_data], [out], quantized=True, input_range=input_range

Instances