Fix pipeline tests - torch imports (#31227)

* Fix pipeline tests - torch imports

* Frameowrk dependant float conversion

src/transformers/pipelines/text_classification.py

@@ -202,7 +202,12 @@ class TextClassificationPipeline(Pipeline):
                 function_to_apply = ClassificationFunction.NONE
 
         outputs = model_outputs["logits"][0]
+
+        if self.framework == "pt":
+            # To enable using fp16 and bf16
             outputs = outputs.float().numpy()
+        else:
+            outputs = outputs.numpy()
 
         if function_to_apply == ClassificationFunction.SIGMOID:
             scores = sigmoid(outputs)

tests/pipelines/test_pipelines_text_classification.py

@@ -14,8 +14,6 @@
 
 import unittest
 
-import torch
-
 from transformers import (
     MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
     TF_MODEL_FOR_SEQUENCE_CLASSIFICATION_MAPPING,
@@ -24,6 +22,7 @@ from transformers import (
 )
 from transformers.testing_utils import (
     is_pipeline_test,
+    is_torch_available,
     nested_simplify,
     require_tf,
     require_torch,
@@ -36,6 +35,10 @@ from transformers.testing_utils import (
 from .test_pipelines_common import ANY
 
 
+if is_torch_available():
+    import torch
+
+
 # These 2 model types require different inputs than those of the usual text models.
 _TO_SKIP = {"LayoutLMv2Config", "LayoutLMv3Config"}