Add back float support to asynchronous_complete_cumsum_cpu (#2383)

Summary: Pull Request resolved: #2383 - Add back float support to `asynchronous_complete_cumsum_cpu` Reviewed By: spcyppt Differential Revision: D54497745 fbshipit-source-id: 3e5f0654aca4414e6256b32f00d97c132fd8e6b0
pytorch · Mar 4, 2024 · d3a6166 · d3a6166
1 parent e764e6f
commit d3a6166
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/fbgemm_gpu/src/sparse_ops/sparse_ops_cpu.cpp b/fbgemm_gpu/src/sparse_ops/sparse_ops_cpu.cpp
@@ -1227,7 +1227,7 @@ Tensor asynchronous_complete_cumsum_cpu(const Tensor& t_in) {
       ? at::empty({t_in.numel() + 1}, t_in.options())
       : at::empty({t_in.size(0), t_in.size(1) + 1}, t_in.options());
 
-  FBGEMM_DISPATCH_INTEGRAL_TYPES(
+  FBGEMM_DISPATCH_ALL_TYPES(
       t_in_contig->scalar_type(),
       "asynchronous_complete_cumsum_cpu_kernel",
       [&] {