diff --git a/src/transformers/testing_utils.py b/src/transformers/testing_utils.py
index e17c5d295fb9..d997bc631b0e 100644
--- a/src/transformers/testing_utils.py
+++ b/src/transformers/testing_utils.py
@@ -165,6 +165,15 @@
 # Not critical, only usable on the sandboxed CI instance.
 TOKEN = "hf_94wBhPGp6KrrTH3KDchhKpRxZwd6dmHWLL"
 
+if is_torch_available():
+    import torch
+
+    IS_ROCM_SYSTEM = torch.version.hip is not None
+    IS_CUDA_SYSTEM = torch.version.cuda is not None
+else:
+    IS_ROCM_SYSTEM = False
+    IS_CUDA_SYSTEM = False
+
 
 def parse_flag_from_env(key, default=False):
     try:
diff --git a/tests/models/bark/test_modeling_bark.py b/tests/models/bark/test_modeling_bark.py
index 476031068f49..bfc6a4dadf9a 100644
--- a/tests/models/bark/test_modeling_bark.py
+++ b/tests/models/bark/test_modeling_bark.py
@@ -1327,4 +1327,9 @@ def test_generate_end_to_end_with_offload(self):
             output_with_offload = self.model.generate(**input_ids, do_sample=False, temperature=1.0)
 
         # checks if same output
-        self.assertListEqual(output_with_no_offload.tolist(), output_with_offload.tolist())
+        self.assertListAlmostEqual(output_with_no_offload.squeeze().tolist(), output_with_offload.squeeze().tolist())
+
+    def assertListAlmostEqual(self, list1, list2, tol=1e-6):
+        self.assertEqual(len(list1), len(list2))
+        for a, b in zip(list1, list2):
+            self.assertAlmostEqual(a, b, delta=tol)
diff --git a/tests/models/gemma/test_modeling_gemma.py b/tests/models/gemma/test_modeling_gemma.py
index 8281f5999371..fa0db9ee7877 100644
--- a/tests/models/gemma/test_modeling_gemma.py
+++ b/tests/models/gemma/test_modeling_gemma.py
@@ -601,6 +601,11 @@ def test_model_2b_fp16_static_cache(self):
     @require_read_token
     def test_model_2b_bf16(self):
         model_id = "google/gemma-2b"
+
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXTS = {
             7: [
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music",
@@ -610,6 +615,10 @@ def test_model_2b_bf16(self):
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music",
                 "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
             ],
+            9: [
+                "Hello I am doing a project on the 1990s and I need to know what the most popular music",
+                "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
+            ],
         }
 
         model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16).to(
@@ -627,6 +636,11 @@ def test_model_2b_bf16(self):
     @require_read_token
     def test_model_2b_eager(self):
         model_id = "google/gemma-2b"
+
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXTS = {
             7: [
                 "Hello I am doing a project on the 1990s and I am looking for some information on the ",
@@ -636,6 +650,10 @@ def test_model_2b_eager(self):
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music",
                 "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
             ],
+            9: [
+                "Hello I am doing a project on the 1990s and I need to know what the most popular music",
+                "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
+            ],
         }
 
         model = AutoModelForCausalLM.from_pretrained(
@@ -655,6 +673,11 @@ def test_model_2b_eager(self):
     @require_read_token
     def test_model_2b_sdpa(self):
         model_id = "google/gemma-2b"
+
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXTS = {
             7: [
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music",
@@ -664,6 +687,10 @@ def test_model_2b_sdpa(self):
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music",
                 "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
             ],
+            9: [
+                "Hello I am doing a project on the 1990s and I need to know what the most popular music",
+                "Hi today I am going to share with you a very easy and simple recipe of <strong><em>Kaju Kat",
+            ],
         }
 
         model = AutoModelForCausalLM.from_pretrained(
@@ -763,6 +790,11 @@ def test_model_7b_fp16(self):
     @require_read_token
     def test_model_7b_bf16(self):
         model_id = "google/gemma-7b"
+
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXTS = {
             7: [
                 """Hello I am doing a project on a 1991 240sx and I am trying to find""",
@@ -772,6 +804,10 @@ def test_model_7b_bf16(self):
                 "Hello I am doing a project for my school and I am trying to make a program that will read a .txt file",
                 "Hi today I am going to show you how to make a very simple and easy to make a very simple and",
             ],
+            9: [
+                "Hello I am doing a project for my school and I am trying to get a servo to move a certain amount of degrees",
+                "Hi today I am going to show you how to make a very simple and easy to make DIY light up sign",
+            ],
         }
 
         model = AutoModelForCausalLM.from_pretrained(model_id, low_cpu_mem_usage=True, torch_dtype=torch.bfloat16).to(
@@ -845,6 +881,11 @@ def test_compile_static_cache(self):
         NUM_TOKENS_TO_GENERATE = 40
         # Note on `EXPECTED_TEXT_COMPLETION`'s diff: the current value matches the original test if the original test
         # was changed to have a cache of 53 tokens (as opposed to 4096), on Ampere GPUs.
+        #
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXT_COMPLETION = {
             8: [
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music was in the 1990s. I have looked on the internet and I have found",
@@ -854,6 +895,10 @@ def test_compile_static_cache(self):
                 "Hello I am doing a project on the 1990s and I need to know what the most popular music was in the 1990s. I have looked on the internet and I have found",
                 "Hi today\nI have a problem with my 2007 1.9 tdi 105bhp.\nI have a problem with the engine management light on.\nI have checked the",
             ],
+            9: [
+                "Hello I am doing a project on the 1990s and I need to know what the most popular music was in the 1990s. I have looked on the internet and I have found",
+                "Hi today\nI have a problem with my 2007 1.9 tdi 105bhp.\nI have a problem with the engine management light on.\nI have checked the",
+            ],
         }
 
         prompts = ["Hello I am doing", "Hi today"]
diff --git a/tests/models/kosmos2/test_modeling_kosmos2.py b/tests/models/kosmos2/test_modeling_kosmos2.py
index 9bc95b8bd44c..ca944d0df034 100644
--- a/tests/models/kosmos2/test_modeling_kosmos2.py
+++ b/tests/models/kosmos2/test_modeling_kosmos2.py
@@ -26,7 +26,7 @@
 
 from transformers import AutoModelForVision2Seq, AutoProcessor, Kosmos2Config
 from transformers.models.kosmos2.configuration_kosmos2 import Kosmos2TextConfig, Kosmos2VisionConfig
-from transformers.testing_utils import require_torch, require_vision, slow, torch_device
+from transformers.testing_utils import IS_ROCM_SYSTEM, require_torch, require_vision, slow, torch_device
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_configuration_common import ConfigTester
@@ -562,6 +562,8 @@ def test_snowman_image_captioning(self):
         processed_text = processed_text[0]
         final_text, entities = final_text_with_entities[0]
 
+        atol = 1e-4 if IS_ROCM_SYSTEM else 1e-5
+
         np.testing.assert_allclose(
             torch.concat(scores[1:4])[:3, :3].to("cpu").numpy(),
             np.array(
@@ -571,7 +573,7 @@ def test_snowman_image_captioning(self):
                     [-0.9352350831031799, -4.688288688659668, 6.240612983703613],
                 ]
             ),
-            atol=1e-5,
+            atol=atol,
         )
         np.testing.assert_allclose(
             torch.concat(scores[-3:])[-3:, -3:].to("cpu").numpy(),
@@ -629,7 +631,7 @@ def test_snowman_image_captioning(self):
                     [-0.7624598741531372, -4.771658897399902, 6.576295852661133],
                 ]
             ),
-            atol=1e-5,
+            atol=atol,
         )
         np.testing.assert_allclose(
             torch.concat(scores[-3:])[-3:, -3:].to("cpu").numpy(),
diff --git a/tests/models/llama/test_modeling_llama.py b/tests/models/llama/test_modeling_llama.py
index 58269d62e08c..61b33b3ec9ab 100644
--- a/tests/models/llama/test_modeling_llama.py
+++ b/tests/models/llama/test_modeling_llama.py
@@ -715,6 +715,11 @@ def test_compile_static_cache(self):
         NUM_TOKENS_TO_GENERATE = 40
         # Note on `EXPECTED_TEXT_COMPLETION`'s diff: the current value matches the original test if the original test
         # was changed to have a cache of 53 tokens (as opposed to 4096), on Ampere GPUs.
+        #
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXT_COMPLETION = {
             8: [
                 "Simply put, the theory of relativity states that 1) the speed of light is constant in all inertial "
@@ -730,7 +735,15 @@ def test_compile_static_cache(self):
                 "My favorite all time favorite condiment is ketchup. I love it on hamburgers, hot dogs, fries, eggs, "
                 "and even on a good old fashioned cheeseburger. I love it on everything. I love it so",
             ],
+            9: [
+                "Simply put, the theory of relativity states that 1) the speed of light is constant in all inertial"
+                " reference frames, and 2) the laws of physics are the same for all inertial reference frames.\nThe "
+                "theory of relativ",
+                "My favorite all time favorite condiment is ketchup. I love it on everything. I love it on my eggs,"
+                " my fries, my chicken, my burgers, my hot dogs, my sandwiches, my salads, my p",
+            ],
         }
+        expected_text_completion_idx = 8
 
         prompts = [
             "Simply put, the theory of relativity states that ",
@@ -745,7 +758,9 @@ def test_compile_static_cache(self):
         # Dynamic Cache
         generated_ids = model.generate(**inputs, max_new_tokens=NUM_TOKENS_TO_GENERATE, do_sample=False)
         dynamic_text = tokenizer.batch_decode(generated_ids, skip_special_tokens=True)
-        self.assertEqual(EXPECTED_TEXT_COMPLETION[8], dynamic_text)  # Both GPU architectures have the same output
+        self.assertEqual(
+            EXPECTED_TEXT_COMPLETION[expected_text_completion_idx], dynamic_text
+        )  # Both GPU architectures have the same output
 
         # Static Cache
         generated_ids = model.generate(
diff --git a/tests/models/mistral/test_modeling_mistral.py b/tests/models/mistral/test_modeling_mistral.py
index dfa0646f8f87..cead89058662 100644
--- a/tests/models/mistral/test_modeling_mistral.py
+++ b/tests/models/mistral/test_modeling_mistral.py
@@ -520,12 +520,16 @@ def test_model_7b_logits(self):
         EXPECTED_MEAN = torch.tensor([[-2.5548, -2.5737, -3.0600, -2.5906, -2.8478, -2.8118, -2.9325, -2.7694]])
         torch.testing.assert_close(out.mean(-1), EXPECTED_MEAN, atol=1e-2, rtol=1e-2)
 
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in output.
         EXPECTED_SLICE = {
             7: torch.tensor([-5.8781, -5.8616, -0.1052, -4.7200, -5.8781, -5.8774, -5.8773, -5.8777, -5.8781, -5.8780, -5.8781, -5.8779, -1.0787, 1.7583, -5.8779, -5.8780, -5.8783, -5.8778, -5.8776, -5.8781, -5.8784, -5.8778, -5.8778, -5.8777, -5.8779, -5.8778, -5.8776, -5.8780, -5.8779, -5.8781]),
             8: torch.tensor([-5.8711, -5.8555, -0.1050, -4.7148, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -1.0781, 1.7568, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711, -5.8711]),
+            9: torch.tensor([-5.8750, -5.8594, -0.1047, -4.7188, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -1.0781,  1.7578, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750, -5.8750]),
         }  # fmt: skip
 
-        print(out[0, 0, :30])
         torch.testing.assert_close(
             out[0, 0, :30], EXPECTED_SLICE[self.cuda_compute_capability_major_version], atol=1e-4, rtol=1e-4
         )
@@ -625,9 +629,14 @@ def test_model_7b_long_prompt_sdpa(self):
 
     @slow
     def test_speculative_generation(self):
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_TEXT_COMPLETION = {
             7: "My favourite condiment is 100% Sriracha. I love the heat, the tang and the fact costs",
             8: "My favourite condiment is 100% ketchup. I love it on everything. I’m not a big",
+            9: "My favourite condiment is 100% ketchup. I love it on everything. I’m not a big",
         }
         prompt = "My favourite condiment is "
         tokenizer = AutoTokenizer.from_pretrained("mistralai/Mistral-7B-v0.1", use_fast=False)
diff --git a/tests/models/mixtral/test_modeling_mixtral.py b/tests/models/mixtral/test_modeling_mixtral.py
index 0972207fce4a..b60062004a4b 100644
--- a/tests/models/mixtral/test_modeling_mixtral.py
+++ b/tests/models/mixtral/test_modeling_mixtral.py
@@ -553,6 +553,10 @@ def test_small_model_logits(self):
         )
         # TODO: might need to tweak it in case the logits do not match on our daily runners
         # these logits have been obtained with the original megablocks impelmentation.
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in output.
         EXPECTED_LOGITS = {
             7: torch.Tensor([[0.1670, 0.1620, 0.6094], [-0.8906, -0.1588, -0.6060], [0.1572, 0.1290, 0.7246]]).to(
                 torch_device
@@ -560,6 +564,9 @@ def test_small_model_logits(self):
             8: torch.Tensor([[0.1631, 0.1621, 0.6094], [-0.8906, -0.1621, -0.6094], [0.1572, 0.1270, 0.7227]]).to(
                 torch_device
             ),
+            9: torch.Tensor([[0.1641, 0.1621, 0.6094], [-0.8906, -0.1631, -0.6094], [0.1572, 0.1260, 0.7227]]).to(
+                torch_device
+            ),
         }
         with torch.no_grad():
             logits = model(dummy_input).logits
@@ -583,6 +590,11 @@ def test_small_model_logits_batched(self):
         )
 
         # TODO: might need to tweak it in case the logits do not match on our daily runners
+        #
+        # Key 9 for MI300, Key 8 for A100/A10, and Key 7 for T4.
+        #
+        # Note: Key 9 is currently set for MI300, but may need potential future adjustments for H100s,
+        # considering differences in hardware processing and potential deviations in generated text.
         EXPECTED_LOGITS_LEFT = {
             7: torch.Tensor(
                 [[0.1750, 0.0537, 0.7007], [0.1750, 0.0537, 0.7007], [0.1750, 0.0537, 0.7007]],
@@ -590,6 +602,9 @@ def test_small_model_logits_batched(self):
             8: torch.Tensor([[0.1914, 0.0508, 0.7188], [0.1953, 0.0510, 0.7227], [0.1973, 0.0562, 0.7148]]).to(
                 torch_device
             ),
+            9: torch.Tensor([[0.1904, 0.0513, 0.7227], [0.1943, 0.0518, 0.7227], [0.1982, 0.0557, 0.7148]]).to(
+                torch_device
+            ),
         }
 
         EXPECTED_LOGITS_LEFT_UNPADDED = {
@@ -599,6 +614,9 @@ def test_small_model_logits_batched(self):
             8: torch.Tensor([[0.2217, 0.5195, -0.3828], [0.8203, -0.2295, 0.6055], [0.2676, -0.7109, 0.2461]]).to(
                 torch_device
             ),
+            9: torch.Tensor([[0.2236, 0.5195, -0.3828], [0.8203, -0.2285, 0.6055], [0.2637, -0.7109, 0.2451]]).to(
+                torch_device
+            ),
         }
 
         EXPECTED_LOGITS_RIGHT_UNPADDED = {
@@ -608,6 +626,9 @@ def test_small_model_logits_batched(self):
             8: torch.Tensor([[0.2178, 0.1260, -0.1621], [-0.3496, 0.2988, -1.0312], [0.0693, 0.7930, 0.8008]]).to(
                 torch_device
             ),
+            9: torch.Tensor([[0.2197, 0.1250, -0.1611], [-0.3516, 0.3008, -1.0312], [0.0684, 0.7930, 0.8008]]).to(
+                torch_device
+            ),
         }
 
         with torch.no_grad():
diff --git a/tests/models/perceiver/test_modeling_perceiver.py b/tests/models/perceiver/test_modeling_perceiver.py
index fbd237bc1058..86bbc7b49d0b 100644
--- a/tests/models/perceiver/test_modeling_perceiver.py
+++ b/tests/models/perceiver/test_modeling_perceiver.py
@@ -26,7 +26,14 @@
 from datasets import load_dataset
 
 from transformers import PerceiverConfig
-from transformers.testing_utils import require_torch, require_torch_multi_gpu, require_vision, slow, torch_device
+from transformers.testing_utils import (
+    IS_ROCM_SYSTEM,
+    require_torch,
+    require_torch_multi_gpu,
+    require_vision,
+    slow,
+    torch_device,
+)
 from transformers.utils import is_torch_available, is_vision_available
 
 from ...test_configuration_common import ConfigTester
@@ -930,7 +937,8 @@ def test_inference_image_classification(self):
 
         expected_slice = torch.tensor([-1.1652, -0.1992, -0.7520], device=torch_device)
 
-        self.assertTrue(torch.allclose(logits[0, :3], expected_slice, atol=1e-4))
+        atol = 1e-3 if IS_ROCM_SYSTEM else 1e-4
+        self.assertTrue(torch.allclose(logits[0, :3], expected_slice, atol=atol))
 
     @slow
     def test_inference_image_classification_fourier(self):
diff --git a/tests/trainer/test_trainer.py b/tests/trainer/test_trainer.py
index 1711f600cebb..239550ac7255 100644
--- a/tests/trainer/test_trainer.py
+++ b/tests/trainer/test_trainer.py
@@ -2507,7 +2507,7 @@ def test_trainer_eval_mrpc(self):
         )
         eval_dataset = GlueDataset(data_args, tokenizer=tokenizer, mode="dev")
 
-        training_args = TrainingArguments(output_dir="./examples", use_cpu=True)
+        training_args = TrainingArguments(output_dir="./examples", use_cpu=True, report_to="none")
         trainer = Trainer(model=model, args=training_args, eval_dataset=eval_dataset)
         result = trainer.evaluate()
         self.assertLess(result["eval_loss"], 0.2)
@@ -2528,6 +2528,7 @@ def test_trainer_eval_multiple(self):
             output_dir="./examples",
             use_cpu=True,
             per_device_eval_batch_size=1,
+            report_to="none",
         )
         trainer = Trainer(
             model=model,
@@ -3063,6 +3064,8 @@ def test_end_to_end_example(self):
                 "--predict_with_generate",
                 "--ddp_timeout",
                 "60",
+                "--report_to",
+                "none",
             ]
             execute_subprocess_async(command)
             # successful return here == success - any errors would have caused an error or a timeout in the sub-call
diff --git a/tests/trainer/test_trainer_seq2seq.py b/tests/trainer/test_trainer_seq2seq.py
index 61d2163b9e81..17593e2e73bf 100644
--- a/tests/trainer/test_trainer_seq2seq.py
+++ b/tests/trainer/test_trainer_seq2seq.py
@@ -161,7 +161,6 @@ def test_return_sequences(self):
             tokenizer=tokenizer,
             data_collator=data_collator,
             compute_metrics=lambda x: {"samples": x[0].shape[0]},
-            report_to="none",
         )
 
         def prepare_data(examples):