[torch_xla2] Fix reenabled op info tests (#8548)

experimental/torch_xla2/test/test_ops.py

@@ -12,14 +12,11 @@
 
 skiplist = {
     "_segment_reduce",
-    "_unsafe_masked_index_put_accumulate",
     "bincount", # NOTE: dtype for int input torch gives float. This is weird.
     "byte",
     "cat",
     "cholesky_solve",
-    "cov",
     "diagonal_copy",
-    "gather",
     "geqrf",
     "histogram", # hard op: AssertionError: Tensor-likes are not close!
     "histogramdd", # TypeError: histogram requires ndarray or scalar arguments, got <class 'list'> at position 1.
@@ -44,7 +41,6 @@
     "normal",
     "ormqr",
     "pca_lowrank",
-    "scatter",
     "searchsorted",
     "special.airy_ai",
     "special.scaled_modified_bessel_k0",
@@ -96,7 +92,8 @@
   'nn.functional.dropout',
 }
 
-atol_dict = {"linalg.eig": (2e0, 3e0),
+atol_dict = {"cov": (2e-1, 2e-4),
+             "linalg.eig": (2e0, 3e0),
              "linalg.eigh": (5e1, 3e0),
              "linalg.eigvalsh": (5e1, 3e0),
              "linalg.pinv": (8e-1, 2e0),

experimental/torch_xla2/torch_xla2/ops/jaten.py

@@ -1745,9 +1745,8 @@ def _aten_atan(self):
 @op(torch.ops.aten.scatter)
 @op(torch.ops.aten.scatter_reduce)
 def _aten_scatter_reduce(input, dim, index, src, reduce, *, include_self=True):
-  if isinstance(src, float):
-    dtype = _torch_binary_scalar_type(src, input)
-    src = jnp.array(src, dtype=dtype)
+  if not isinstance(src, jnp.ndarray):
+    src = jnp.array(src, dtype=input.dtype)
   input_indexes, source_indexes = _scatter_index(dim, index)
   # "Zero out" target elements when not included
   if not include_self:
@@ -2596,6 +2595,9 @@ def _aten_frexp(input):
 def _aten_gather(input, dim, index):
   if input.ndim == 0:
     return jnp.broadcast_to(input, index.shape)
+  # short circuit for empty outputs
+  if not all(index.shape):
+    return jnp.zeros(index.shape, dtype=input.dtype)
   if dim < 0:
     dim += input.ndim
   input_indexes, source_indexes = _scatter_index(dim, index)
@@ -4732,9 +4734,9 @@ def _new_empty_strided(self, size, stride, dtype=None, **kwargs):
     return jnp.empty(size, dtype=jax_dtype)
 
 
-@op(torch.ops.aten._unsafe_index_put, is_jax_function=False)
+@op(torch.ops.aten._unsafe_index_put)
 def _aten_unsafe_index_put(self, indices, values, accumulate=False):
-  return self.index_put_(indices, values, accumulate)
+  return _aten_index_put(self, indices, values, accumulate)
 
 
 @op(torch.ops.aten.conj_physical,