Update prosody_embedding_pipeline.py

prosody_embedding_pipeline.py

@@ -5,7 +5,6 @@ import torch
 from typing import Dict, Union, List, Optional
 from pathlib import Path
 import logging
-from .prosody_preprocessor import ProsodyPreprocessor, ProsodyConfig
 from datasets import Dataset
 
 logger = logging.getLogger(__name__)
@@ -16,46 +15,15 @@ class ProsodyEmbeddingPipeline(Pipeline):
         speaker_stats,
         f0_interp,
         f0_normalize,
-        preprocessor: Optional[ProsodyPreprocessor] = None,
         stats_dir: Optional[str] = None,
         **kwargs
     ):
         super().__init__(**kwargs)
-        self.preprocessor = preprocessor or ProsodyPreprocessor()
         self.stats_dir = Path(stats_dir) if stats_dir else None
         self.speaker_stats = speaker_stats
         self.f0_interp = f0_interp
         self.f0_normalize = f0_normalize
         
-
-    @classmethod
-    def from_dataset(
-        cls,
-        dataset: Dataset,
-        stats_dir: str = "preprocessor_stats",
-        **kwargs
-    ) -> "ProsodyPipeline":
-        """Initialize pipeline by computing speaker statistics from a dataset"""
-        logger.info("Initializing pipeline from dataset...")
-        
-        preprocessor = ProsodyPreprocessor()
-        
-        stats_dir = Path(stats_dir)
-        stats_dir.mkdir(parents=True, exist_ok=True)
-        
-        logger.info("Computing speaker statistics...")
-        features_dataset, speaker_stats = preprocessor.collect_stats(dataset)
-        
-        stats_path = stats_dir / "speaker_stats.pt"
-        logger.info(f"Saving speaker statistics to {stats_path}")
-        preprocessor.save_stats(stats_path)
-        
-        return cls(
-            preprocessor=preprocessor,
-            stats_dir=stats_dir,
-            model=None,
-            **kwargs
-        )
     
 
     def _sanitize_parameters(self, **kwargs):
@@ -229,23 +197,3 @@ class ProsodyEmbeddingPipeline(Pipeline):
             
         return outputs
 
-
-    @classmethod
-    def from_pretrained(cls, save_directory: Union[str, Path], **kwargs):
-        """Load a pretrained pipeline"""
-        save_directory = Path(save_directory)
-        
-        config = ProsodyConfig.from_pretrained(save_directory)
-        preprocessor = ProsodyPreprocessor(config)
-        
-        pipeline = cls(
-            preprocessor=preprocessor,
-            stats_dir=save_directory,
-            **kwargs
-        )
-        
-        stats_path = save_directory / "speaker_stats.pt"
-        if stats_path.exists():
-            pipeline.speaker_stats = ProsodyPreprocessor.load_stats(stats_path)
-        
-        return pipeline