diff --git a/tests/models/granitemoe/test_modeling_granitemoe.py b/tests/models/granitemoe/test_modeling_granitemoe.py
index 0c2e0b5e5eab..609b53052f6f 100644
--- a/tests/models/granitemoe/test_modeling_granitemoe.py
+++ b/tests/models/granitemoe/test_modeling_granitemoe.py
@@ -298,12 +298,12 @@ class GraniteMoeModelTest(ModelTesterMixin, GenerationTesterMixin, unittest.Test
     # This is because we are hitting edge cases with the causal_mask buffer
     model_split_percents = [0.5, 0.7, 0.8]
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch=['gfx1201','gfx942'])
     def test_generate_with_static_cache(self):
         super().test_generate_with_static_cache()
         pass
 
-    @skipIfRocm(arch='gfx1201')
+    @skipIfRocm(arch=['gfx1201','gfx942'])
     def test_generate_from_inputs_embeds_with_static_cache(self):
         super().test_generate_from_inputs_embeds_with_static_cache()
         pass
diff --git a/tests/models/hubert/test_modeling_hubert.py b/tests/models/hubert/test_modeling_hubert.py
index 25ed24a892ef..ae6202ea119d 100644
--- a/tests/models/hubert/test_modeling_hubert.py
+++ b/tests/models/hubert/test_modeling_hubert.py
@@ -590,7 +590,7 @@ def test_model(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_model(*config_and_inputs)
 
-    @skipIfRocm(arch='gfx90a')
+    @skipIfRocm(arch=['gfx90a','gfx942'])
     def test_batched_inference(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_batch_inference(*config_and_inputs)
diff --git a/tests/models/mamba2/test_modeling_mamba2.py b/tests/models/mamba2/test_modeling_mamba2.py
index c5b787f64389..e52284d45739 100644
--- a/tests/models/mamba2/test_modeling_mamba2.py
+++ b/tests/models/mamba2/test_modeling_mamba2.py
@@ -20,7 +20,7 @@
 from parameterized import parameterized
 
 from transformers import AutoTokenizer, Mamba2Config, is_torch_available
-from transformers.testing_utils import require_read_token, require_torch, require_torch_gpu, slow, torch_device
+from transformers.testing_utils import require_read_token, require_torch, require_torch_gpu, slow, torch_device, skipIfRocm
 from transformers.utils.import_utils import is_causal_conv1d_available, is_mamba_2_ssm_available
 
 from ...generation.test_utils import GenerationTesterMixin
@@ -233,6 +233,7 @@ def setUp(self):
             self, config_class=Mamba2Config, n_embd=37, common_properties=["hidden_size", "num_hidden_layers"]
         )
 
+    @skipIfRocm(arch='gfx942')
     def test_mamba2_caching(self):
         config_and_inputs = self.model_tester.prepare_config_and_inputs()
         self.model_tester.create_and_check_mamba2_caching(*config_and_inputs)
diff --git a/tests/models/speech_encoder_decoder/test_modeling_speech_encoder_decoder.py b/tests/models/speech_encoder_decoder/test_modeling_speech_encoder_decoder.py
index 897d4b056f19..75d0ca8a1bed 100644
--- a/tests/models/speech_encoder_decoder/test_modeling_speech_encoder_decoder.py
+++ b/tests/models/speech_encoder_decoder/test_modeling_speech_encoder_decoder.py
@@ -24,6 +24,7 @@
     require_torch_sdpa,
     slow,
     torch_device,
+    skipIfRocm,
 )
 
 from ...test_modeling_common import floats_tensor, ids_tensor, random_attention_mask
@@ -503,6 +504,10 @@ def test_sdpa_can_dispatch_composite_models(self):
 
 @require_torch
 class Wav2Vec2BertModelTest(EncoderDecoderMixin, unittest.TestCase):
+    @skipIfRocm(arch='gfx942')
+    def test_save_and_load_from_pretrained(self):
+        super().test_save_and_load_from_pretrained()
+
     def get_pretrained_model_and_inputs(self):
         model = SpeechEncoderDecoderModel.from_encoder_decoder_pretrained(
             "facebook/wav2vec2-base-960h", "google-bert/bert-base-cased"