remove unnecessary entries in some auto model mappings (#45224)

fixxxxxx

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

Author: ydshieh

src/transformers/models/auto/modeling_auto.py

@@ -1009,7 +1009,6 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         ("perception_lm", "PerceptionLMForConditionalGeneration"),
         ("pi0", "PI0ForConditionalGeneration"),
         ("pix2struct", "Pix2StructForConditionalGeneration"),
-        ("pixtral", "LlavaForConditionalGeneration"),
         ("pp_chart2table", "GotOcr2ForConditionalGeneration"),
         ("qwen2_5_vl", "Qwen2_5_VLForConditionalGeneration"),
         ("qwen2_vl", "Qwen2VLForConditionalGeneration"),
@@ -1671,7 +1670,6 @@ class _BaseModelWithGenerate(PreTrainedModel, GenerationMixin):
         # Model for Text-To-Waveform mapping
         ("bark", "BarkModel"),
         ("csm", "CsmForConditionalGeneration"),
-        ("fastspeech2_conformer", "FastSpeech2ConformerWithHifiGan"),
         ("fastspeech2_conformer_with_hifigan", "FastSpeech2ConformerWithHifiGan"),
         ("higgs_audio_v2", "HiggsAudioV2ForConditionalGeneration"),
         ("musicgen", "MusicgenForConditionalGeneration"),

src/transformers/models/fastspeech2_conformer/configuration_fastspeech2_conformer.py

@@ -262,7 +262,7 @@ def validate_architecture(self):
             raise ValueError("Either use_masking or use_weighted_masking can be True, but not both.")
 
 
-@auto_docstring(checkpoint="espnet/fastspeech2_conformer")
+@auto_docstring(checkpoint="espnet/fastspeech2_conformer_with_hifigan")
 @strict
 class FastSpeech2ConformerHifiGanConfig(PreTrainedConfig):
     r"""
@@ -323,7 +323,7 @@ def __post_init__(self, **kwargs):
         super().__post_init__(**kwargs)
 
 
-@auto_docstring(checkpoint="espnet/fastspeech2_conformer")
+@auto_docstring(checkpoint="espnet/fastspeech2_conformer_with_hifigan")
 @strict
 class FastSpeech2ConformerWithHifiGanConfig(PreTrainedConfig):
     r"""