fix tf.roll with axes=-1 (#1338)

Signed-off-by: Guenther Schmuelling <guschmue@microsoft.com>

tests/test_backend.py

@@ -911,6 +911,15 @@ def func(x, shift):
             return tf.identity(x_, name=_TFOUTPUT)
         self._run_test_case(func, [_OUTPUT], {_INPUT: x_val, _INPUT1: shift_val})
 
+    @check_opset_min_version(10, "Slice")
+    def test_roll_neg_axis(self):
+        def func(input_ids):
+            shifted_input_ids = tf.cast(input_ids, tf.int32)
+            shifted_input_ids = tf.roll(shifted_input_ids, 1, axis=-1)
+            return tf.identity(shifted_input_ids, name=_TFOUTPUT)
+        x_val = np.array([[0, 1, 2, 3, 4, 5, 6, 7], [1, 2, 3, 4, 5, 6, 7, 8]], dtype=np.int64)
+        self._run_test_case(func, [_OUTPUT], {_INPUT: x_val})
+
     @check_tf_min_version("2.2")
     def test_large_model_format(self):
         x_val = np.array([2.0], dtype=np.float32)
@@ -4581,5 +4590,6 @@ def func(x):
         self._run_test_case(func, [_OUTPUT], {_INPUT: x_val})
 
 
+
 if __name__ == '__main__':
     unittest_main()

tf2onnx/onnx_opset/tensor.py

@@ -356,6 +356,8 @@ class Roll:
     def any_version(cls, opset, ctx, node, **kwargs):
         utils.make_sure(node.inputs[2].is_const(), "Can only convert Roll is axis is const")
         axes = node.inputs[2].get_tensor_value()
+        if axes == -1:
+            axes = len(ctx.get_shape(node.input[0])) + axes
         if not isinstance(axes, list):
             axes = [axes]
         shifts_dtype = ctx.get_dtype(node.input[1])