🚨🚨🚨 Remove softmax for EfficientNetForImageClassification 🚨🚨🚨 (#25501)

* Remove softmax for EfficientNet

* Update integration test values

* Fix up

src/transformers/models/efficientnet/modeling_efficientnet.py

@@ -588,7 +588,6 @@ class EfficientNetForImageClassification(EfficientNetPreTrainedModel):
         # Classifier head
         self.dropout = nn.Dropout(p=config.dropout_rate)
         self.classifier = nn.Linear(config.hidden_dim, self.num_labels) if self.num_labels > 0 else nn.Identity()
-        self.classifier_act = nn.Softmax(dim=1)
 
         # Initialize weights and apply final processing
         self.post_init()
@@ -620,7 +619,6 @@ class EfficientNetForImageClassification(EfficientNetPreTrainedModel):
         pooled_output = outputs.pooler_output if return_dict else outputs[1]
         pooled_output = self.dropout(pooled_output)
         logits = self.classifier(pooled_output)
-        logits = self.classifier_act(logits)
 
         loss = None
         if labels is not None:

tests/models/efficientnet/test_modeling_efficientnet.py

@@ -265,5 +265,5 @@ class EfficientNetModelIntegrationTest(unittest.TestCase):
         expected_shape = torch.Size((1, 1000))
         self.assertEqual(outputs.logits.shape, expected_shape)
 
-        expected_slice = torch.tensor([0.0001, 0.0002, 0.0002]).to(torch_device)
+        expected_slice = torch.tensor([-0.2962, 0.4487, 0.4499]).to(torch_device)
         self.assertTrue(torch.allclose(outputs.logits[0, :3], expected_slice, atol=1e-4))