Remove ConversationalPipeline and Conversation object (#31165)

* Remove ConversationalPipeline and Conversation object, as they have been deprecated for some time and are due for removal

* Update not-doctested.txt

* Fix JA and ZH docs

* Fix JA and ZH docs some more

* Fix JA and ZH docs some more

tests/models/led/test_modeling_led.py

@@ -284,7 +284,6 @@ class LEDModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     all_generative_model_classes = (LEDForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": LEDForConditionalGeneration,
             "feature-extraction": LEDModel,
             "question-answering": LEDForQuestionAnswering,
             "summarization": LEDForConditionalGeneration,

tests/models/led/test_modeling_tf_led.py

@@ -197,7 +197,6 @@ class TFLEDModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestCase)
     all_generative_model_classes = (TFLEDForConditionalGeneration,) if is_tf_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": TFLEDForConditionalGeneration,
             "feature-extraction": TFLEDModel,
             "summarization": TFLEDForConditionalGeneration,
             "text2text-generation": TFLEDForConditionalGeneration,

tests/models/longt5/test_modeling_longt5.py

@@ -504,7 +504,6 @@ class LongT5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     all_generative_model_classes = (LongT5ForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": LongT5ForConditionalGeneration,
             "feature-extraction": LongT5Model,
             "summarization": LongT5ForConditionalGeneration,
             "text2text-generation": LongT5ForConditionalGeneration,

tests/models/m2m_100/test_modeling_m2m_100.py

@@ -243,7 +243,6 @@ class M2M100ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     all_generative_model_classes = (M2M100ForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": M2M100ForConditionalGeneration,
             "feature-extraction": M2M100Model,
             "summarization": M2M100ForConditionalGeneration,
             "text2text-generation": M2M100ForConditionalGeneration,

tests/models/marian/test_modeling_flax_marian.py

@@ -311,10 +311,6 @@ class FlaxMarianModelTest(FlaxModelTesterMixin, unittest.TestCase, FlaxGeneratio
             outputs = model(input_ids)
             self.assertIsNotNone(outputs)
 
-    @unittest.skip("Skipping for now, to fix @ArthurZ or @ydshieh")
-    def test_pipeline_conversational(self):
-        pass
-
 
 @require_flax
 @require_sentencepiece

tests/models/marian/test_modeling_marian.py

@@ -248,7 +248,6 @@ class MarianModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     all_generative_model_classes = (MarianMTModel,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": MarianMTModel,
             "feature-extraction": MarianModel,
             "summarization": MarianMTModel,
             "text-generation": MarianForCausalLM,
@@ -350,10 +349,6 @@ class MarianModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     def test_tie_word_embeddings_decoder(self):
         pass
 
-    @unittest.skip("Skipping for now, to fix @ArthurZ or @ydshieh")
-    def test_pipeline_conversational(self):
-        pass
-
     @unittest.skip(
         reason="This architecure seem to not compute gradients properly when using GC, check: https://github.com/huggingface/transformers/pull/27124"
     )

tests/models/marian/test_modeling_tf_marian.py

@@ -184,7 +184,6 @@ class TFMarianModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestCa
     all_generative_model_classes = (TFMarianMTModel,) if is_tf_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": TFMarianMTModel,
             "feature-extraction": TFMarianModel,
             "summarization": TFMarianMTModel,
             "text2text-generation": TFMarianMTModel,
@@ -208,10 +207,6 @@ class TFMarianModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestCa
         config_and_inputs = self.model_tester.prepare_config_and_inputs_for_common()
         self.model_tester.check_decoder_model_past_large_inputs(*config_and_inputs)
 
-    @unittest.skip("Skipping for now, to fix @ArthurZ or @ydshieh")
-    def test_pipeline_conversational(self):
-        pass
-
 
 @require_tf
 class AbstractMarianIntegrationTest(unittest.TestCase):

tests/models/mbart/test_modeling_mbart.py

@@ -240,7 +240,6 @@ class MBartModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixi
     all_generative_model_classes = (MBartForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": MBartForConditionalGeneration,
             "feature-extraction": MBartModel,
             "fill-mask": MBartForConditionalGeneration,
             "question-answering": MBartForQuestionAnswering,

tests/models/mbart/test_modeling_tf_mbart.py

@@ -161,7 +161,6 @@ class TFMBartModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestCas
     all_generative_model_classes = (TFMBartForConditionalGeneration,) if is_tf_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": TFMBartForConditionalGeneration,
             "feature-extraction": TFMBartModel,
             "summarization": TFMBartForConditionalGeneration,
             "text2text-generation": TFMBartForConditionalGeneration,

tests/models/mt5/test_modeling_mt5.py

@@ -555,7 +555,6 @@ class MT5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     all_generative_model_classes = (MT5ForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": MT5ForConditionalGeneration,
             "feature-extraction": MT5Model,
             "question-answering": MT5ForQuestionAnswering,
             "summarization": MT5ForConditionalGeneration,
@@ -886,10 +885,6 @@ class MT5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
             attn_weights = out[attn_name] if attn_name == attention_names[0] else out[attn_name][-1]
             self.assertEqual(sum([w.sum().item() for w in attn_weights]), 0.0)
 
-    @unittest.skip("Does not support conversations.")
-    def test_pipeline_conversational(self):
-        pass
-
 
 # Copied from tests.models.t5.test_modeling_t5.T5EncoderOnlyModelTester with T5->MT5
 class MT5EncoderOnlyModelTester:

tests/models/mvp/test_modeling_mvp.py

@@ -421,7 +421,6 @@ class MvpModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     all_generative_model_classes = (MvpForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": MvpForConditionalGeneration,
             "feature-extraction": MvpModel,
             "fill-mask": MvpForConditionalGeneration,
             "question-answering": MvpForQuestionAnswering,

tests/models/nllb_moe/test_modeling_nllb_moe.py

@@ -250,7 +250,6 @@ class NllbMoeModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     all_generative_model_classes = (NllbMoeForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": NllbMoeForConditionalGeneration,
             "feature-extraction": NllbMoeModel,
             "summarization": NllbMoeForConditionalGeneration,
             "text2text-generation": NllbMoeForConditionalGeneration,

tests/models/pegasus/test_modeling_pegasus.py

@@ -246,7 +246,6 @@ class PegasusModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMi
     all_generative_model_classes = (PegasusForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": PegasusForConditionalGeneration,
             "feature-extraction": PegasusModel,
             "summarization": PegasusForConditionalGeneration,
             "text-generation": PegasusForCausalLM,

tests/models/pegasus/test_modeling_tf_pegasus.py

@@ -182,7 +182,6 @@ class TFPegasusModelTest(TFModelTesterMixin, PipelineTesterMixin, unittest.TestC
     all_generative_model_classes = (TFPegasusForConditionalGeneration,) if is_tf_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": TFPegasusForConditionalGeneration,
             "feature-extraction": TFPegasusModel,
             "summarization": TFPegasusForConditionalGeneration,
             "text2text-generation": TFPegasusForConditionalGeneration,

tests/models/pegasus_x/test_modeling_pegasus_x.py

@@ -206,7 +206,6 @@ class PegasusXModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterM
     all_generative_model_classes = (PegasusXForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": PegasusXForConditionalGeneration,
             "feature-extraction": PegasusXModel,
             "summarization": PegasusXForConditionalGeneration,
             "text2text-generation": PegasusXForConditionalGeneration,

tests/models/plbart/test_modeling_plbart.py

@@ -227,7 +227,6 @@ class PLBartModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMix
     all_generative_model_classes = (PLBartForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": PLBartForConditionalGeneration,
             "feature-extraction": PLBartModel,
             "summarization": PLBartForConditionalGeneration,
             "text-classification": PLBartForSequenceClassification,

tests/models/prophetnet/test_modeling_prophetnet.py

@@ -891,7 +891,6 @@ class ProphetNetModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTeste
     all_generative_model_classes = (ProphetNetForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": ProphetNetForConditionalGeneration,
             "feature-extraction": ProphetNetModel,
             "summarization": ProphetNetForConditionalGeneration,
             "text-generation": ProphetNetForCausalLM,

tests/models/seamless_m4t/test_modeling_seamless_m4t.py

@@ -645,7 +645,6 @@ class SeamlessM4TModelWithTextInputTest(
     pipeline_model_mapping = (
         {
             "automatic-speech-recognition": SeamlessM4TForSpeechToText,
-            "conversational": SeamlessM4TForTextToText,
             "feature-extraction": SeamlessM4TModel,
             "summarization": SeamlessM4TForTextToText,
             "text-to-audio": SeamlessM4TForTextToSpeech,

tests/models/switch_transformers/test_modeling_switch_transformers.py

@@ -559,7 +559,6 @@ class SwitchTransformersModelTest(ModelTesterMixin, GenerationTesterMixin, Pipel
     all_generative_model_classes = (SwitchTransformersForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": SwitchTransformersForConditionalGeneration,
             "feature-extraction": SwitchTransformersModel,
             "summarization": SwitchTransformersForConditionalGeneration,
             "text2text-generation": SwitchTransformersForConditionalGeneration,

tests/models/t5/test_modeling_t5.py

@@ -558,7 +558,6 @@ class T5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
     all_generative_model_classes = (T5ForConditionalGeneration,) if is_torch_available() else ()
     pipeline_model_mapping = (
         {
-            "conversational": T5ForConditionalGeneration,
             "feature-extraction": T5Model,
             "question-answering": T5ForQuestionAnswering,
             "summarization": T5ForConditionalGeneration,
@@ -889,10 +888,6 @@ class T5ModelTest(ModelTesterMixin, GenerationTesterMixin, PipelineTesterMixin,
             attn_weights = out[attn_name] if attn_name == attention_names[0] else out[attn_name][-1]
             self.assertEqual(sum([w.sum().item() for w in attn_weights]), 0.0)
 
-    @unittest.skip("Does not support conversations.")
-    def test_pipeline_conversational(self):
-        pass
-
 
 class T5EncoderOnlyModelTester:
     def __init__(