run peephole to do profile-based optimizations (pytorch#33337)

Summary: We need to run a peephole before constant propagation in the profiling pipeline, so we fold `prim::shape` for inputs with complete tensor types. Pull Request resolved: pytorch#33337 Differential Revision: D19905624 Pulled By: Krovatkin fbshipit-source-id: 80fff067941556053847ddc7afe0fd1c7a89a3ba
gaurav1086 · Feb 20, 2020 · 36d724c · 36d724c
1 parent 1a25747
commit 36d724c
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/torch/csrc/jit/profiling_graph_executor_impl.cpp b/torch/csrc/jit/profiling_graph_executor_impl.cpp
@@ -84,6 +84,7 @@ void ProfilingGraphExecutorImpl::runProfilingOptimizations(
   specializeAutogradZero(*copy);
 
   runRequiredPasses(copy);
+  PeepholeOptimize(copy);
   ConstantPropagation(copy);
   runOptimization(copy);