diffs/36712.patch

diff --git a/src/transformers/models/blenderbot/modeling_blenderbot.py b/src/transformers/models/blenderbot/modeling_blenderbot.py
index 16bea0a09f41..e40208a65b65 100755
--- a/src/transformers/models/blenderbot/modeling_blenderbot.py
+++ b/src/transformers/models/blenderbot/modeling_blenderbot.py
@@ -320,9 +320,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/m2m_100/modeling_m2m_100.py b/src/transformers/models/m2m_100/modeling_m2m_100.py
index 63a387911827..e16a299dfcd2 100755
--- a/src/transformers/models/m2m_100/modeling_m2m_100.py
+++ b/src/transformers/models/m2m_100/modeling_m2m_100.py
@@ -631,9 +631,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/mbart/modeling_mbart.py b/src/transformers/models/mbart/modeling_mbart.py
index 17d2af3c13d1..c1cd512f85b6 100755
--- a/src/transformers/models/mbart/modeling_mbart.py
+++ b/src/transformers/models/mbart/modeling_mbart.py
@@ -580,9 +580,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/pegasus/modeling_pegasus.py b/src/transformers/models/pegasus/modeling_pegasus.py
index fe7008f1612f..32e105053bcb 100755
--- a/src/transformers/models/pegasus/modeling_pegasus.py
+++ b/src/transformers/models/pegasus/modeling_pegasus.py
@@ -321,9 +321,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/qwen2_audio/modeling_qwen2_audio.py b/src/transformers/models/qwen2_audio/modeling_qwen2_audio.py
index c5d45c9c9d1f..8a9122c48708 100644
--- a/src/transformers/models/qwen2_audio/modeling_qwen2_audio.py
+++ b/src/transformers/models/qwen2_audio/modeling_qwen2_audio.py
@@ -427,9 +427,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/speech_to_text/modeling_speech_to_text.py b/src/transformers/models/speech_to_text/modeling_speech_to_text.py
index ccff216c98c1..fd52e64a1065 100755
--- a/src/transformers/models/speech_to_text/modeling_speech_to_text.py
+++ b/src/transformers/models/speech_to_text/modeling_speech_to_text.py
@@ -386,9 +386,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)
 
diff --git a/src/transformers/models/whisper/modeling_whisper.py b/src/transformers/models/whisper/modeling_whisper.py
index be0465d6c63c..5f7cf59e8e9f 100644
--- a/src/transformers/models/whisper/modeling_whisper.py
+++ b/src/transformers/models/whisper/modeling_whisper.py
@@ -637,9 +637,7 @@ def forward(
         hidden_states = nn.functional.dropout(hidden_states, p=self.dropout, training=self.training)
         hidden_states = residual + hidden_states
 
-        if hidden_states.dtype == torch.float16 and (
-            torch.isinf(hidden_states).any() or torch.isnan(hidden_states).any()
-        ):
+        if hidden_states.dtype == torch.float16:
             clamp_value = torch.finfo(hidden_states.dtype).max - 1000
             hidden_states = torch.clamp(hidden_states, min=-clamp_value, max=clamp_value)