From 6b5b7b81341a032e11691aae92b27d48c0390169 Mon Sep 17 00:00:00 2001
From: Meghan Cowan <mcowan@nvidia.com>
Date: Tue, 19 Nov 2024 18:23:23 -0800
Subject: [PATCH] Skip sequence parallel test when D=1 (#3451)

Add gtest skip when for distributed transformer sequence parallel when
number of devices=1.
---
 tests/cpp/test_multidevice_transformer.cpp | 6 ++++++
 1 file changed, 6 insertions(+)

diff --git a/tests/cpp/test_multidevice_transformer.cpp b/tests/cpp/test_multidevice_transformer.cpp
index 2f4bda3f77e..2ef33dcdf8f 100644
--- a/tests/cpp/test_multidevice_transformer.cpp
+++ b/tests/cpp/test_multidevice_transformer.cpp
@@ -886,6 +886,9 @@ TEST_P(DistributedTransformerTest, MultiheadAttention_SP) {
     GTEST_SKIP() << "Requires number of devices=" << D
                  << " evenly divide H=" << H;
   }
+  if (D == 1) {
+    GTEST_SKIP() << "Requires >1 devices, D=" << D;
+  }
   auto dtype = GetParam();
   auto fusion = std::make_unique<Fusion>();
   FusionGuard fg(fusion.get());
@@ -1125,6 +1128,9 @@ TEST_P(DistributedTransformerTest, Forward_SP) {
     GTEST_SKIP() << "Requires number of devices=" << D
                  << " evenly divide H=" << H;
   }
+  if (D == 1) {
+    GTEST_SKIP() << "Requires >1 devices, D=" << D;
+  }
   auto dtype = GetParam();
   at::ScalarType at_dtype = data_type_to_aten(dtype);
   auto fusion = std::make_unique<Fusion>();