[ExecuTorch][WebGPU] Add squeeze_copy + unsqueeze_copy (flat copies)

Pull Request resolved: #20392

Adds `aten.squeeze_copy.dims` and `aten.unsqueeze_copy.default` to the WebGPU delegate. Both are numel-preserving shape ops; on a dense row-major buffer backend they are the same flat copy as `view_copy` — only the shape metadata differs (mirrors the Vulkan delegate, which routes both through `add_view_copy_node`).

Composition (no new kernel):
- `squeeze/Squeeze.cpp` — reads `args = [self, dims, out]`, ignores the AOT-fixed `dims`, calls `add_flat_copy(graph, in, out)` from `runtime/ops/view_copy/view_copy.h`.
- `unsqueeze/Unsqueeze.cpp` — reads `args = [self, dim, out]`, ignores the AOT-fixed `dim`, calls `add_flat_copy(graph, in, out)`.
ghstack-source-id: 397026523
@exported-using-ghexport

Differential Revision: [D108793153](https://our.internmc.facebook.com/intern/diff/D108793153/)

Author: JulianCloudNTH

backends/webgpu/CMakeLists.txt

@@ -45,6 +45,8 @@ set(WEBGPU_SRCS
     runtime/ops/view_copy/ViewCopy.cpp
     runtime/ops/select/Select.cpp
     runtime/ops/sigmoid/UnaryOp.cpp
+    runtime/ops/squeeze/Squeeze.cpp
+    runtime/ops/unsqueeze/Unsqueeze.cpp
 )
 
 add_library(webgpu_backend ${WEBGPU_SRCS})

backends/webgpu/runtime/ops/squeeze/Squeeze.cpp

@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#include <executorch/backends/webgpu/runtime/WebGPUGraph.h>
+#include <executorch/backends/webgpu/runtime/ops/OperatorRegistry.h>
+#include <executorch/backends/webgpu/runtime/ops/view_copy/view_copy.h>
+
+#include <vector>
+
+namespace executorch::backends::webgpu {
+
+namespace {
+
+// squeeze_copy.dims = numel-preserving flat copy (Vulkan Squeeze.cpp:102-104).
+void squeeze_copy_dims_impl(WebGPUGraph& graph, const std::vector<int>& args) {
+  // args: [self, dims, out]; dims ignored (out shape fixed AOT).
+  add_flat_copy(graph, args.at(0), args.at(args.size() - 1));
+}
+
+} // namespace
+
+WEBGPU_REGISTER_OPERATORS {
+  WEBGPU_REGISTER_OP(aten.squeeze_copy.dims, squeeze_copy_dims_impl);
+}
+
+} // namespace executorch::backends::webgpu

backends/webgpu/runtime/ops/unsqueeze/Unsqueeze.cpp

@@ -0,0 +1,31 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#include <executorch/backends/webgpu/runtime/WebGPUGraph.h>
+#include <executorch/backends/webgpu/runtime/ops/OperatorRegistry.h>
+#include <executorch/backends/webgpu/runtime/ops/view_copy/view_copy.h>
+
+#include <vector>
+
+namespace executorch::backends::webgpu {
+
+namespace {
+
+// unsqueeze_copy = numel-preserving flat copy (Vulkan Unsqueeze.cpp:101-103).
+void unsqueeze_copy_impl(WebGPUGraph& graph, const std::vector<int>& args) {
+  // args: [self, dim, out]; dim ignored (out shape fixed AOT, like view_copy).
+  add_flat_copy(graph, args.at(0), args.at(args.size() - 1));
+}
+
+} // namespace
+
+WEBGPU_REGISTER_OPERATORS {
+  WEBGPU_REGISTER_OP(aten.unsqueeze_copy.default, unsqueeze_copy_impl);
+}
+
+} // namespace executorch::backends::webgpu