From a09daacf6bc9c304d4492044dcd38ab6fd7d2789 Mon Sep 17 00:00:00 2001
From: root <root@banff-cyxtera-s80-2.amd.com>
Date: Fri, 16 Aug 2024 20:02:56 +0000
Subject: [PATCH] [release/2.3] skip test_autocast_rnn

---
 test/test_cuda.py | 1 +
 1 file changed, 1 insertion(+)

diff --git a/test/test_cuda.py b/test/test_cuda.py
index cefb72c587b1..5f10f2441665 100644
--- a/test/test_cuda.py
+++ b/test/test_cuda.py
@@ -1636,6 +1636,7 @@ def forward(self):
     # cudnn RNNs require special backend handling (weights are cast to FP16 and reflattened)
     # so they get a dedicated test.
     # Despite the large number of RNN cases it tries, the test takes < 15 seconds on a Titan V (similar to V100).
+    @skipIfRocm
     @unittest.skipIf(not TEST_CUDNN, 'CUDNN not available')
     def test_autocast_rnn(self):
         with torch.backends.cudnn.flags(enabled=True, deterministic=True):