[ONNX][TOPI][RELAY] Resize refactor (#7883)

* adds rounding mode for nearest neighbor, passing onnx unit tests for nearest neighbor

* passing all linear test. passing all nearest tests except crop and resize, which needs a dynamic implementation of crop and resize

* most of the bicubic tests are working

* fix exclude outside

* remove dead code

* fix lint

* fix defaults to match old implementation

* fix lint

* fix gpu tests

* fix lint again

* change order of operations to prevent GPU rounding errors

include/tvm/relay/attrs/image.h

@@ -38,6 +38,9 @@ struct ResizeAttrs : public tvm::AttrsNode<ResizeAttrs> {
   std::string layout;
   std::string method;
   std::string coordinate_transformation_mode;
+  std::string rounding_method;
+  double bicubic_alpha;
+  int bicubic_exclude;
   DataType out_dtype;
 
   TVM_DECLARE_ATTRS(ResizeAttrs, "relay.attrs.ResizeAttrs") {
@@ -61,6 +64,17 @@ struct ResizeAttrs : public tvm::AttrsNode<ResizeAttrs> {
             "to the coordinate in the original tensor."
             "Refer to the ONNX Resize operator specification for details"
             "Available options are half_pixel, align_corners and asymmetric");
+    TVM_ATTR_FIELD(rounding_method)
+        .set_default("round")
+        .describe(
+            "indicates how to find the \"nearest\" pixel in nearest_neighbor method"
+            "Available options are round, floor, and ceil.");
+    TVM_ATTR_FIELD(bicubic_alpha)
+        .set_default(-0.5)
+        .describe("Spline Coefficient for Bicubic Interpolation");
+    TVM_ATTR_FIELD(bicubic_exclude)
+        .set_default(0)
+        .describe("Flag to exclude exterior of the image during bicubic interpolation");
     TVM_ATTR_FIELD(out_dtype).set_default(NullValue<DataType>()).describe("Output data type.");
   }
 };

python/tvm/relay/frontend/onnx.py

@@ -2088,11 +2088,13 @@ class Resize(OnnxOpConverter):
 
     @classmethod
     def _impl_v10(cls, inputs, attr, params):
-        mode = attr.get("mode")
-        if mode == b"nearest":
+        mode = attr.get("mode").decode("ascii")
+        if mode == "nearest":
             method = "nearest_neighbor"
-        elif mode == b"linear":
+        elif mode == "linear":
             method = "bilinear"
+        elif mode == "cubic":
+            method = "bicubic"
         else:
             raise tvm.error.OpAttributeInvalid(
                 'Value {} in attribute "mode" of operator Resize is not valid.'.format(mode)
@@ -2106,16 +2108,25 @@ def _impl_v10(cls, inputs, attr, params):
 
     @classmethod
     def _impl_v11(cls, inputs, attr, params):
-        mode = attr.get("mode")
-        if mode == b"nearest":
+        layout = "NCHW"  # ONNX assumes NCHW layout
+
+        mode = attr.get("mode").decode("ascii")
+        if mode == "nearest":
             method = "nearest_neighbor"
-        elif mode == b"linear":
+        elif mode == "linear":
             method = "bilinear"
+        elif mode == "cubic":
+            method = "bicubic"
         else:
             raise tvm.error.OpAttributeInvalid(
                 'Value {} in attribute "mode" of operator Resize is not valid.'.format(mode)
             )
 
+        coord_trans = attr.get("coordinate_transformation_mode", b"half_pixel").decode("ascii")
+        nearest_mode = attr.get("nearest_mode", b"round_prefer_floor").decode("ascii")
+        alpha = attr.get("cubic_coeff_a", -0.75)
+        exclude = attr.get("exclude_outside", 0)
+
         scale = inputs[2]
         scale_shape = infer_shape(scale)
         if len(inputs) == 4:
@@ -2126,21 +2137,11 @@ def _impl_v11(cls, inputs, attr, params):
         else:
             assert len(scale_shape) != 0, "One of scale or size should be passed."
             size = _op.cast(shape_of(inputs[0]), infer_type(scale).checked_type.dtype) * scale
-
-        coord_trans = attr.get("coordinate_transformation_mode")
-        if coord_trans in [b"pytorch_half_pixel", b"half_pixel"]:
-            coord_trans = "half_pixel"
-        elif coord_trans == b"align_corners":
-            coord_trans = "align_corners"
-        elif coord_trans == b"asymmetric" or method == "nearest_neighbor":
-            coord_trans = "asymmetric"
-        else:
-            raise tvm.error.OpAttributeInvalid(
-                "Unsupported coordinate_transformation_mode: {}".format(coord_trans)
-            )
-        layout = "NCHW"  # ONNX assumes NCHW layout
         out_size = fold_constant(_op.strided_slice(size, [2], [4]))
-        return _op.image.resize(inputs[0], out_size, layout, method, coord_trans)
+
+        return _op.image.resize(
+            inputs[0], out_size, layout, method, coord_trans, nearest_mode, alpha, exclude
+        )
 
 
 class NonZero(OnnxOpConverter):

python/tvm/relay/op/dyn/image/_image.py

@@ -31,10 +31,22 @@ def compute_resize(attrs, inputs, out_type):
     layout = attrs.layout
     method = attrs.method
     coord_trans = attrs.coordinate_transformation_mode
+    rounding_method = attrs.rounding_method
+    bicubic_alpha = attrs.bicubic_alpha
+    bicubic_exclude = attrs.bicubic_exclude
     out_dtype = attrs.out_dtype
     return [
         tvm.topi.image.resize(
-            inputs[0], inputs[1], layout, method, coord_trans, out_dtype, out_type.shape
+            inputs[0],
+            inputs[1],
+            layout,
+            method,
+            coord_trans,
+            rounding_method,
+            bicubic_alpha,
+            bicubic_exclude,
+            out_dtype,
+            out_type.shape,
         )
     ]
 

python/tvm/relay/op/image/_image.py

@@ -31,12 +31,28 @@
 # resize
 @reg.register_compute("image.resize")
 def compute_resize(attrs, inputs, out_type):
+    """ compute definition for resize op """
     size = attrs.size
     layout = attrs.layout
     method = attrs.method
     coord_trans = attrs.coordinate_transformation_mode
+    rounding_method = attrs.rounding_method
+    bicubic_alpha = attrs.bicubic_alpha
+    bicubic_exclude = attrs.bicubic_exclude
     out_dtype = attrs.out_dtype
-    return [topi.image.resize(inputs[0], size, layout, method, coord_trans, out_dtype)]
+    return [
+        topi.image.resize(
+            inputs[0],
+            size,
+            layout,
+            method,
+            coord_trans,
+            rounding_method,
+            bicubic_alpha,
+            bicubic_exclude,
+            out_dtype,
+        )
+    ]
 
 
 reg.register_injective_schedule("image.resize")

python/tvm/relay/op/image/image.py

@@ -26,6 +26,9 @@ def resize(
     layout="NCHW",
     method="bilinear",
     coordinate_transformation_mode="half_pixel",
+    rounding_method="",
+    bicubic_alpha=-0.5,
+    bicubic_exclude=0,
     out_dtype=None,
 ):
     """Image resize operator.
@@ -58,6 +61,16 @@ def resize(
         Refer to the ONNX Resize operator specification for details.
         [half_pixel, align_corners, asymmetric]
 
+    rounding_method: string, optional
+        indicates how to find the "nearest" pixel in nearest_neighbor method
+        [round, floor, ceil]
+
+    bicubic_alpha: float
+        Spline Coefficient for Bicubic Interpolation
+
+    bicubic_exclude: int
+            Flag to exclude exterior of the image during bicubic interpolation
+
     out_dtype : str, optional
         Type to return. If left None returns the same type as input.
 
@@ -70,9 +83,27 @@ def resize(
         size = list(size.data.asnumpy().astype("int32"))
     if isinstance(size, Expr):
         return _dyn_make.resize(
-            data, size, layout, method, coordinate_transformation_mode, out_dtype
+            data,
+            size,
+            layout,
+            method,
+            coordinate_transformation_mode,
+            rounding_method,
+            bicubic_alpha,
+            bicubic_exclude,
+            out_dtype,
         )
-    return _make.resize(data, size, layout, method, coordinate_transformation_mode, out_dtype)
+    return _make.resize(
+        data,
+        size,
+        layout,
+        method,
+        coordinate_transformation_mode,
+        rounding_method,
+        bicubic_alpha,
+        bicubic_exclude,
+        out_dtype,
+    )
 
 
 def resize3d(
@@ -151,7 +182,7 @@ def crop_and_resize(
         A 1-D tensor of shape [num_boxes], box_ind[i] specifies the data that
         the i-th box refers to.
 
-    crop_size : Tuple of Expr
+    crop_size : Tuple of PrimExpr
         The target size to which each box will be resized.
 
     layout : str, optional