From f2ecb9eec42d9f55bd4190f3755f63dab816a083 Mon Sep 17 00:00:00 2001
From: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Date: Mon, 17 Oct 2022 11:56:29 -0400
Subject: [PATCH] Revert "add return_tensor parameter for feature extraction
 (#19257)" (#19680)

This reverts commit 35bd089a241788a43a43e27de1ef3f5cede7954b.
---
 .../pipelines/feature_extraction.py            | 14 +++-----------
 .../test_pipelines_feature_extraction.py       | 18 ------------------
 2 files changed, 3 insertions(+), 29 deletions(-)

diff --git a/src/transformers/pipelines/feature_extraction.py b/src/transformers/pipelines/feature_extraction.py
index 48f7735b6ce0b4..c7d50c971955fd 100644
--- a/src/transformers/pipelines/feature_extraction.py
+++ b/src/transformers/pipelines/feature_extraction.py
@@ -31,8 +31,6 @@ 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*):
@@ -42,7 +40,7 @@ class FeatureExtractionPipeline(Pipeline):
             the associated CUDA device id.
     """
 
-    def _sanitize_parameters(self, truncation=None, tokenize_kwargs=None, return_tensors=None, **kwargs):
+    def _sanitize_parameters(self, truncation=None, tokenize_kwargs=None, **kwargs):
         if tokenize_kwargs is None:
             tokenize_kwargs = {}
 
@@ -55,11 +53,7 @@ def _sanitize_parameters(self, truncation=None, tokenize_kwargs=None, return_ten
 
         preprocess_params = tokenize_kwargs
 
-        postprocess_params = {}
-        if return_tensors is not None:
-            postprocess_params["return_tensors"] = return_tensors
-
-        return preprocess_params, {}, postprocess_params
+        return preprocess_params, {}, {}
 
     def preprocess(self, inputs, **tokenize_kwargs) -> Dict[str, GenericTensor]:
         return_tensors = self.framework
@@ -70,10 +64,8 @@ def _forward(self, model_inputs):
         model_outputs = self.model(**model_inputs)
         return model_outputs
 
-    def postprocess(self, model_outputs, return_tensors=False):
+    def postprocess(self, model_outputs):
         # [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":
diff --git a/tests/pipelines/test_pipelines_feature_extraction.py b/tests/pipelines/test_pipelines_feature_extraction.py
index 2e431fa1d48612..f75af6808bccdf 100644
--- a/tests/pipelines/test_pipelines_feature_extraction.py
+++ b/tests/pipelines/test_pipelines_feature_extraction.py
@@ -15,8 +15,6 @@
 import unittest
 
 import numpy as np
-import tensorflow as tf
-import torch
 
 from transformers import (
     FEATURE_EXTRACTOR_MAPPING,
@@ -135,22 +133,6 @@ def test_tokenization_small_model_tf(self):
                 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 = []