diff --git a/torch_patches/.torch_pin b/torch_patches/.torch_pin
new file mode 100644
index 000000000000..06eab6aef1a3
--- /dev/null
+++ b/torch_patches/.torch_pin
@@ -0,0 +1 @@
+#120076
diff --git a/torch_xla/csrc/aten_xla_type.cpp b/torch_xla/csrc/aten_xla_type.cpp
index 36791d0fc056..e9733a204c4d 100644
--- a/torch_xla/csrc/aten_xla_type.cpp
+++ b/torch_xla/csrc/aten_xla_type.cpp
@@ -829,7 +829,7 @@ at::Tensor XLANativeFunctions::baddbmm(const at::Tensor& self,
 }
 
 at::Tensor XLANativeFunctions::bernoulli(
-    const at::Tensor& self, c10::optional<at::Generator> generator) {
+    const at::Tensor& self, const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,
@@ -841,7 +841,7 @@ at::Tensor XLANativeFunctions::bernoulli(
 }
 
 at::Tensor XLANativeFunctions::bernoulli(
-    const at::Tensor& self, double p, c10::optional<at::Generator> generator) {
+    const at::Tensor& self, double p, const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -853,7 +853,7 @@ at::Tensor XLANativeFunctions::bernoulli(
 
 at::Tensor& XLANativeFunctions::bernoulli_(
     at::Tensor& self, const at::Tensor& p,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -1244,7 +1244,7 @@ at::Tensor XLANativeFunctions::expand_copy_symint(const at::Tensor& self,
 }
 
 at::Tensor& XLANativeFunctions::exponential_(
-    at::Tensor& self, double lambd, c10::optional<at::Generator> generator) {
+    at::Tensor& self, double lambd, const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,
@@ -1927,7 +1927,7 @@ at::Tensor XLANativeFunctions::mul(const at::Tensor& self,
 
 at::Tensor XLANativeFunctions::multinomial(
     const at::Tensor& self, int64_t num_samples, bool replacement,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   XLA_CHECK(num_samples > 0)
       << "Multinomial number of samples must be greater than 0";
   XLA_CHECK(at::isFloatingType(self.scalar_type()))
@@ -2242,7 +2242,7 @@ at::Tensor XLANativeFunctions::norm(const at::Tensor& self,
 }
 
 at::Tensor XLANativeFunctions::normal(const at::Tensor& mean, double std,
-                                      c10::optional<at::Generator> generator) {
+                                      const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -2254,7 +2254,7 @@ at::Tensor XLANativeFunctions::normal(const at::Tensor& mean, double std,
 }
 
 at::Tensor XLANativeFunctions::normal(double mean, const at::Tensor& std,
-                                      c10::optional<at::Generator> generator) {
+                                      const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -2267,7 +2267,7 @@ at::Tensor XLANativeFunctions::normal(double mean, const at::Tensor& std,
 
 at::Tensor XLANativeFunctions::normal(const at::Tensor& mean,
                                       const at::Tensor& std,
-                                      c10::optional<at::Generator> generator) {
+                                      const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -2280,7 +2280,7 @@ at::Tensor XLANativeFunctions::normal(const at::Tensor& mean,
 
 at::Tensor& XLANativeFunctions::normal_(
     at::Tensor& self, double mean, double std,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,
@@ -2417,7 +2417,7 @@ std::tuple<at::Tensor, at::Tensor> XLANativeFunctions::qr(
 // The value generated should be within (from, to].
 at::Tensor& XLANativeFunctions::random_(
     at::Tensor& self, int64_t from, c10::optional<int64_t> to,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<
@@ -2437,7 +2437,7 @@ at::Tensor& XLANativeFunctions::random_(
 
 // The value generated should be in (0, to].
 at::Tensor& XLANativeFunctions::random_(
-    at::Tensor& self, int64_t to, c10::optional<at::Generator> generator) {
+    at::Tensor& self, int64_t to, const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,
@@ -2453,7 +2453,7 @@ at::Tensor& XLANativeFunctions::random_(
 
 // The value generated should be in (self_type_min, self_type_max).
 at::Tensor& XLANativeFunctions::random_(
-    at::Tensor& self, c10::optional<at::Generator> generator) {
+    at::Tensor& self, const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,
@@ -2594,7 +2594,7 @@ at::Tensor XLANativeFunctions::roll(const at::Tensor& self,
 at::Tensor XLANativeFunctions::rrelu_with_noise(
     const at::Tensor& self, const at::Tensor& noise, const at::Scalar& lower,
     const at::Scalar& upper, bool training,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     // The fallback path for rrelu_with_noise when training=true is wrong
@@ -3133,7 +3133,7 @@ std::vector<at::Tensor> XLANativeFunctions::unbind_copy(const at::Tensor& self,
 
 at::Tensor& XLANativeFunctions::uniform_(
     at::Tensor& self, double from, double to,
-    c10::optional<at::Generator> generator) {
+    const std::optional<at::Generator>& generator) {
   TORCH_LAZY_FN_COUNTER_TIMED_TRACING("xla::");
   if (generator.has_value() && generator->defined()) {
     return at::native::call_fallback_fn<&xla_cpu_fallback,