fix whisper re-compile (#36712)

* fix whisper re-compile

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix copy

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix comment

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* fix copies

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

* revert useless changes

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>

---------

Signed-off-by: jiqing-feng <jiqing.feng@intel.com>
Co-authored-by: Marc Sun <57196510+SunMarc@users.noreply.github.com>

src/transformers/models/blenderbot/modeling_blenderbot.py

@@ -320,9 +320,7 @@ class BlenderbotEncoderLayer(nn.Module):
         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)
 

src/transformers/models/m2m_100/modeling_m2m_100.py

@@ -631,9 +631,7 @@ class M2M100EncoderLayer(nn.Module):
         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)
 

src/transformers/models/mbart/modeling_mbart.py

@@ -580,9 +580,7 @@ class MBartEncoderLayer(nn.Module):
         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)
 

src/transformers/models/pegasus/modeling_pegasus.py

@@ -321,9 +321,7 @@ class PegasusEncoderLayer(nn.Module):
         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)
 

src/transformers/models/qwen2_audio/modeling_qwen2_audio.py

@@ -427,9 +427,7 @@ class Qwen2AudioEncoderLayer(nn.Module):
         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)
 

src/transformers/models/speech_to_text/modeling_speech_to_text.py

@@ -386,9 +386,7 @@ class Speech2TextEncoderLayer(nn.Module):
         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)
 

src/transformers/models/whisper/modeling_whisper.py

@@ -637,9 +637,7 @@ class WhisperEncoderLayer(nn.Module):
         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)