add return_tensor parameter for feature extraction (#19257)

* add return_tensors parameter for feature_extraction  w/ test

add return_tensor parameter for feature extraction

Revert "Merge branch 'feature-extraction-return-tensor' of https://github.com/ajsanjoaquin/transformers into feature-extraction-return-tensor"

This reverts commit d559da743b87914e111a84a98ba6dbb70d08ad88, reversing
changes made to bbef89278650c04c090beb65637a8e9572dba222.

* call parameter directly

Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>

* Fixup.

* Update src/transformers/pipelines/feature_extraction.py

Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

Co-authored-by: Nicolas Patry <patry.nicolas@protonmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

src/transformers/pipelines/feature_extraction.py

@@ -31,6 +31,8 @@ class FeatureExtractionPipeline(Pipeline):
             If no framework is specified, will default to the one currently installed. If no framework is specified and
             both frameworks are installed, will default to the framework of the `model`, or to PyTorch if no model is
             provided.
+        return_tensor (`bool`, *optional*):
+            If `True`, returns a tensor according to the specified framework, otherwise returns a list.
         task (`str`, defaults to `""`):
             A task-identifier for the pipeline.
         args_parser ([`~pipelines.ArgumentHandler`], *optional*):
@@ -40,7 +42,7 @@ class FeatureExtractionPipeline(Pipeline):
             the associated CUDA device id.
     """
 
-    def _sanitize_parameters(self, truncation=None, tokenize_kwargs=None, **kwargs):
+    def _sanitize_parameters(self, truncation=None, tokenize_kwargs=None, return_tensors=None, **kwargs):
         if tokenize_kwargs is None:
             tokenize_kwargs = {}
 
@@ -53,7 +55,11 @@ class FeatureExtractionPipeline(Pipeline):
 
         preprocess_params = tokenize_kwargs
 
-        return preprocess_params, {}, {}
+        postprocess_params = {}
+        if return_tensors is not None:
+            postprocess_params["return_tensors"] = return_tensors
+
+        return preprocess_params, {}, postprocess_params
 
     def preprocess(self, inputs, **tokenize_kwargs) -> Dict[str, GenericTensor]:
         return_tensors = self.framework
@@ -64,8 +70,10 @@ class FeatureExtractionPipeline(Pipeline):
         model_outputs = self.model(**model_inputs)
         return model_outputs
 
-    def postprocess(self, model_outputs):
+    def postprocess(self, model_outputs, return_tensors=False):
         # [0] is the first available tensor, logits or last_hidden_state.
+        if return_tensors:
+            return model_outputs[0]
         if self.framework == "pt":
             return model_outputs[0].tolist()
         elif self.framework == "tf":

tests/pipelines/test_pipelines_feature_extraction.py

@@ -15,6 +15,8 @@
 import unittest
 
 import numpy as np
+import tensorflow as tf
+import torch
 
 from transformers import (
     FEATURE_EXTRACTOR_MAPPING,
@@ -133,6 +135,22 @@ class FeatureExtractionPipelineTests(unittest.TestCase, metaclass=PipelineTestCa
                 tokenize_kwargs=tokenize_kwargs,
             )
 
+    @require_torch
+    def test_return_tensors_pt(self):
+        feature_extractor = pipeline(
+            task="feature-extraction", model="hf-internal-testing/tiny-random-distilbert", framework="pt"
+        )
+        outputs = feature_extractor("This is a test" * 100, return_tensors=True)
+        self.assertTrue(torch.is_tensor(outputs))
+
+    @require_tf
+    def test_return_tensors_tf(self):
+        feature_extractor = pipeline(
+            task="feature-extraction", model="hf-internal-testing/tiny-random-distilbert", framework="tf"
+        )
+        outputs = feature_extractor("This is a test" * 100, return_tensors=True)
+        self.assertTrue(tf.is_tensor(outputs))
+
     def get_shape(self, input_, shape=None):
         if shape is None:
             shape = []