Implement upsample_bilinear2d_aa

experimental/torch_xla2/test/test_ops.py

@@ -14,7 +14,6 @@
     "__rpow__",  # NOTE: cannot fix because torch test case has undefined behavior
                  # such as 0 to negative power.
     "_segment_reduce",
-    "_upsample_bilinear2d_aa",
     "bincount", # NOTE: dtype for int input torch gives float. This is weird.
     "byte",
     "cat",
@@ -270,6 +269,7 @@ def replace_values_below_threshold(self, torch_tensor, threshold):
   def test_reference_eager(self, device, dtype, op):
     sample_inputs = op.sample_inputs(device, dtype)
     for sample_input in sample_inputs:
+      print("sample_input: ", op.name, sample_input)
       t = sample_input.input
       if isinstance(t, torch.Tensor) and t.is_sparse:
         continue

experimental/torch_xla2/torch_xla2/ops/jaten.py

@@ -1,9 +1,10 @@
 """Torch ops implemented using jax."""
 
 import sys
-from typing import Optional, Sequence
+from typing import Optional, Sequence, Tuple, Union
 import functools
 
+import math
 import jax
 from jax import numpy as jnp
 import functools
@@ -4095,3 +4096,59 @@ def _aten_max_unpoolxd(input, indices, output_size, stride=None, padding=0):
 
     return output
 
+@op(torch.ops.aten._upsample_bilinear2d_aa)
+def _aten_upsample_bilinear2d_aa(input, output_size, align_corners, scale_factors):
+    # input: is of type jaxlib.xla_extension.ArrayImpl
+    image = input
+
+    # https://jax.readthedocs.io/en/latest/_autosummary/jax.image.resize.html
+    # Resize does not distinguish batch, channel size.
+    # We need to leave them as is
+    # https://pytorch.org/vision/stable/transforms.html#supported-input-types-and-conventions
+    # pytorch image shape is (C,H,W) or (N,C,H,W)
+    # N - batch size
+    # C - no of channels
+    # H,W - heigth, width
+
+    shape = list(image.shape)
+    # overriding output_size
+    if output_size:
+      shape[-1] = output_size[-1]
+      shape[-2] = output_size[-2]
+    if scale_factors:
+      shape[-1] = int(math.floor(shape[-1]*scale_factors[-1]))
+      shape[-2] = int(math.floor(shape[-2]*scale_factors[-2]))
+    method = "bilinear"
+    antialias = True # ignored for upsampling
+
+    # align_corners is not supported in resize()
+    # https://github.com/jax-ml/jax/issues/11206
+    if align_corners:
+      return resize_with_aligned_corners(image, shape, method, antialias=True)
+    return jax.image.resize(image, shape, method, antialias) # precision=Precision.HIGHEST
+
+# From: https://github.com/jax-ml/jax/issues/11206
+def resize_with_aligned_corners(
+    image: jax.Array,
+    shape: Tuple[int, ...],
+    method: Union[str, jax.image.ResizeMethod],
+    antialias: bool,
+):
+    """Alternative to jax.image.resize(), which emulates align_corners=True in PyTorch's
+    interpolation functions."""
+    spatial_dims = tuple(
+        i
+        for i in range(len(shape))
+        if not jax.core.symbolic_equal_dim(image.shape[i], shape[i])
+    )
+    scale = jnp.array([(shape[i] - 1.0) / (image.shape[i] - 1.0) for i in spatial_dims])
+    translation = -(scale / 2.0 - 0.5)
+    return jax.image.scale_and_translate(
+        image,
+        shape,
+        method=method,
+        scale=scale,
+        spatial_dims=spatial_dims,
+        translation=translation,
+        antialias=antialias,
+    )