make_seq_tensor in codegen (#3088)

Summary:
Pull Request resolved: #3088

increasing sequence is very useful for development, particularly for "slicing" and "indexing" operations.
ghstack-source-id: 222827546

Reviewed By: SS-JIA

Differential Revision: D56095314

fbshipit-source-id: 1491bb2399581eb303472b572c74c070c833d654

Author: yipjustin

backends/vulkan/test/op_tests/utils/codegen_base.py

@@ -218,6 +218,25 @@ def generate_suite_cpp(self) -> str:
     return at::rand(sizes, at::device(at::kCPU).dtype(dtype)) * (high - low) + low;
 }}
 
+
+at::Tensor make_seq_tensor(
+    std::vector<int64_t> sizes,
+      at::ScalarType dtype = at::kFloat) {{
+  int64_t n = 1;
+  for (auto size: sizes) {{
+    n *= size;
+  }}
+
+  std::vector<float> values(n);
+  for (int i=0;i<n;i++) {{
+    values[i] = (float) i;
+  }}
+
+  // from_blob doesn't take ownership of data. Hence must create a copy as
+  // "values" will go out of scope.
+  return at::from_blob(values.data(), sizes, dtype).detach().clone();
+}}
+
 {test_suites_cpp}
 """