diff --git a/caffe2/python/recurrent.py b/caffe2/python/recurrent.py
index d4762f08c683e3..8bb0d9cfd6d656 100644
--- a/caffe2/python/recurrent.py
+++ b/caffe2/python/recurrent.py
@@ -282,7 +282,7 @@ def map_to_dual_list(m):
     cell_net.Proto().type = 'simple'
 
     # The last output is a list of step workspaces,
-    # which is only needed internally for gradient propogation
+    # which is only needed internally for gradient propagation
     return results[:-1]
 
 
diff --git a/test/distributed/test_c10d_nccl.py b/test/distributed/test_c10d_nccl.py
index b68765013e76d5..f41390157cd1d5 100644
--- a/test/distributed/test_c10d_nccl.py
+++ b/test/distributed/test_c10d_nccl.py
@@ -2183,7 +2183,7 @@ def div(fut):
             process_group, allreduce_with_then_hook
         )
 
-        # check whether the grads are equal to what allreduce returns multuplied by 5.
+        # check whether the grads are equal to what allreduce returns multiplied by 5.
         # without the comm_hook, result would be still 0.25 * torch.ones(2, 2).
         self._run_and_verify_hook(gpu_model, 8, 1.25 * torch.ones(2, 2))
 
diff --git a/test/test_fx.py b/test/test_fx.py
index 8fb7c76aa28e83..3e3b12fb671480 100644
--- a/test/test_fx.py
+++ b/test/test_fx.py
@@ -1550,7 +1550,7 @@ def forward(self, x):
         self.assertEqual(opcodes, set(['placeholder', 'get_attr', 'call_function', 'call_method',
                                        'call_module', 'output']))
 
-        # Test shape propogation and make sure results match actual
+        # Test shape propagation and make sure results match actual
         self.assertEqual(output_shape, ref_out.shape)
         self.assertEqual(output_stride, ref_out.stride())
 
diff --git a/torch/csrc/jit/runtime/autodiff.cpp b/torch/csrc/jit/runtime/autodiff.cpp
index a8dc3cc7859a7d..7e1f6182534cc2 100644
--- a/torch/csrc/jit/runtime/autodiff.cpp
+++ b/torch/csrc/jit/runtime/autodiff.cpp
@@ -389,7 +389,7 @@ bool outputRequiresGrad(Value* output) {
 static ReverseDetails addReverseInline(Gradient& grad_desc) {
   auto& graph = *grad_desc.f;
   // note: reverse_node is intentionally not inserted to avoid
-  // accidentally acting on it (e.g. in elminate dead code),
+  // accidentally acting on it (e.g. in eliminate dead code),
   // std::cout << *reverse_node << to view its state.
   auto reverse_node = graph.create(prim::Reverse, 0);
   auto reverse_block = reverse_node->addBlock();
diff --git a/torchgen/gen.py b/torchgen/gen.py
index 2ff7773562da56..807fa092371abe 100644
--- a/torchgen/gen.py
+++ b/torchgen/gen.py
@@ -103,7 +103,7 @@
 #   - 'api' has conversions for how to translate JIT schema into
 #     the various C++ APIs that the codegen interacts with.  There
 #     are in fact THREE different C++ APIs: the public C++ API,
-#     the dispatcher API, and the legacy disaptcher API.  See each
+#     the dispatcher API, and the legacy dispatcher API.  See each
 #     of these respective files for more information
 
 # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ #