Update on "make etrecord support export program"

Differential Revision: [D77965102](https://our.internmc.facebook.com/intern/diff/D77965102/)

[ghstack-poisoned]

Author: Gasoonjia

.ci/docker/ci_commit_pins/pytorch.txt

@@ -1 +1 @@
-7cda4017ddda554752e89069ae205be5e8388f59
+9b498d3bb28b8e3411ce464dd2755c5b96d92c8f

.ci/scripts/check_c10_sync.sh

@@ -12,4 +12,4 @@ pushd pytorch
 git checkout "$pytorch_pin"
 popd
 "$(dirname "${BASH_SOURCE[0]}")"/compare_dirs.sh runtime/core/portable_type/c10/c10 pytorch/c10
-"$(dirname "${BASH_SOURCE[0]}")"/compare_dirs.sh runtime/core/portable_type/c10/torch/headeronly pytorch/torch/headeronly
+"$(dirname "${BASH_SOURCE[0]}")"/compare_dirs.sh runtime/core/portable_type/c10/torch/standalone pytorch/torch/standalone

backends/arm/_passes/arm_pass_manager.py

@@ -6,6 +6,8 @@
 # LICENSE file in the root directory of this source tree.
 
 # pyre-unsafe
+
+import executorch.backends.arm.tosa.dialect  # noqa: unused
 from executorch.backends.arm._passes import (
     AddBiasPass,
     AnnotateChannelsLastDimOrder,

backends/arm/test/passes/test_decorate_fp32_to_int32_casting_pass.py

@@ -6,7 +6,7 @@
 from typing import Tuple
 
 import torch
-from executorch.backends.arm.test import common
+from executorch.backends.arm.test import common, conftest
 
 from executorch.backends.arm.test.tester.test_pipeline import (
     OpNotSupportedPipeline,
@@ -55,6 +55,7 @@ def test_decorate_fp32_to_int32_casting_tosa_MI(test_data: Tuple):
         (test_tensor,),
         aten_op=[],
         exir_op=[],
+        run_on_tosa_ref_model=conftest.is_option_enabled("tosa_ref_model"),
     )
     pipeline.run()
 

backends/cadence/fusion_g3/operators/op_exp.cpp

@@ -60,7 +60,7 @@ Tensor& exp_out(KernelRuntimeContext& ctx, const Tensor& in, Tensor& out) {
     return out;
   } else {
     return torch::executor::native::internal::
-        unary_ufunc_realhbbf16_to_floathbf16(std::exp, ctx, in, out);
+        unary_ufunc_realhbbf16_to_floathbf16(std::exp, std::exp, ctx, in, out);
   }
 }
 

backends/cadence/fusion_g3/operators/op_rsqrt.cpp

@@ -27,7 +27,8 @@ namespace native {
 
 namespace {
 
-double rsqrt(double x) {
+template <typename T>
+T rsqrt(T x) {
   return 1.0 / std::sqrt(x);
 }
 
@@ -61,11 +62,11 @@ Tensor& rsqrt_out(KernelRuntimeContext& ctx, const Tensor& in, Tensor& out) {
     return out;
   } else {
     return torch::executor::native::internal::
-        unary_ufunc_realhbbf16_to_floathbf16(rsqrt, ctx, in, out);
+        unary_ufunc_realhbbf16_to_floathbf16(rsqrt, rsqrt, ctx, in, out);
   }
 }
 
 } // namespace native
 } // namespace G3
 } // namespace impl
-} // namespace cadence
+} // namespace cadence

backends/cadence/fusion_g3/operators/op_sqrt.cpp

@@ -55,7 +55,8 @@ Tensor& sqrt_out(KernelRuntimeContext& ctx, const Tensor& in, Tensor& out) {
     return out;
   } else {
     return torch::executor::native::internal::
-        unary_ufunc_realhbbf16_to_floathbf16(std::sqrt, ctx, in, out);
+        unary_ufunc_realhbbf16_to_floathbf16(
+            std::sqrt, std::sqrt, ctx, in, out);
   }
 }
 

backends/cadence/fusion_g3/operators/op_tanh.cpp

@@ -55,7 +55,8 @@ Tensor& tanh_out(KernelRuntimeContext& ctx, const Tensor& in, Tensor& out) {
     return out;
   } else {
     return torch::executor::native::internal::
-        unary_ufunc_realhbbf16_to_floathbf16(std::tanh, ctx, in, out);
+        unary_ufunc_realhbbf16_to_floathbf16(
+            std::tanh, std::tanh, ctx, in, out);
   }
 }
 

backends/cadence/hifi/operators/op_rsqrt.cpp

@@ -21,7 +21,8 @@ namespace HiFi {
 namespace native {
 namespace {
 
-double rsqrt(double x) {
+template <typename T>
+T rsqrt(T x) {
   return 1.0 / std::sqrt(x);
 }
 
@@ -46,7 +47,7 @@ Tensor& rsqrt_out(RuntimeContext& ctx, const Tensor& in, Tensor& out) {
   }
 
   return torch::executor::native::internal::
-      unary_ufunc_realhbbf16_to_floathbf16(rsqrt, ctx, in, out);
+      unary_ufunc_realhbbf16_to_floathbf16(rsqrt, rsqrt, ctx, in, out);
 }
 
 } // namespace native

backends/cadence/hifi/operators/op_tanh.cpp

@@ -35,10 +35,10 @@ Tensor& tanh_out(RuntimeContext& ctx, const Tensor& in, Tensor& out) {
   }
 
   return torch::executor::native::internal::
-      unary_ufunc_realhbbf16_to_floathbf16(std::tanh, ctx, in, out);
+      unary_ufunc_realhbbf16_to_floathbf16(std::tanh, std::tanh, ctx, in, out);
 }
 
 } // namespace native
 } // namespace HiFi
 } // namespace impl
-} // namespace cadence
+} // namespace cadence