debugging 2

laughter-detection/transcribe.py

@@ -5,10 +5,15 @@ import torch
 from pydub import AudioSegment
 from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
 from tqdm import tqdm
+import logging
+
+# Setup logging
+logging.basicConfig(level=logging.DEBUG, format='%(asctime)s - %(levelname)s - %(message)s')
 
 def create_directory(path):
     if not os.path.exists(path):
         os.makedirs(path)
+    logging.debug(f"Directory created at: {path}")
 
 def transcribe_audio(file_path, pipe, transcripts_dir):
     try:
@@ -19,13 +24,13 @@ def transcribe_audio(file_path, pipe, transcripts_dir):
             txt_file_path = os.path.join(transcripts_dir, txt_file_name)
             with open(txt_file_path, 'w') as txt_file:
                 txt_file.write('\n'.join(str(chunk) for chunk in result["chunks"]))
-            print(f"Transcription saved to {txt_file_path}")
+            logging.info(f"Transcription saved to {txt_file_path}")
             return txt_file_path
         else:
-            print("No transcription was generated.")
+            logging.warning("No transcription was generated.")
             return None
     except ValueError as e:
-        print(f"Error processing file : {e}")
+        logging.error(f"Error processing file {file_path} : {e}")
         return None
 
 def check_timestamps_and_slice(audio_file_path, transcript_file_path, chunks_dir, jump_threshold, timestamp_records):
@@ -37,11 +42,11 @@ def check_timestamps_and_slice(audio_file_path, transcript_file_path, chunks_dir
     for i in range(len(rows) - 1):
         current_end_time = rows[i]['timestamp'][1] * 1000 if rows[i]['timestamp'][1] is not None else None
         next_start_time = rows[i + 1]['timestamp'][0] * 1000 if rows[i+1]['timestamp'][1] is not None else None
-        if current_end_time is not None and next_start_time is not None:
-            if next_start_time - current_end_time > jump_threshold:
-                timestamps.append((int(current_end_time), int(next_start_time)))
+        if current_end_time is not None and next_start_time is not None and (next_start_time - current_end_time > jump_threshold):
+            timestamps.append((int(current_end_time), int(next_start_time)))
+            logging.debug(f"Timestamp slice between {current_end_time} and {next_start_time} identified.")
         else:
-            print(f"Skipping segment due to missing timestamp: current_end_time={current_end_time}, next_start_time={next_start_time}")
+            logging.debug(f"Skipping segment due to missing or insufficient gap in timestamp: current_end_time={current_end_time}, next_start_time={next_start_time}")
 
     episode_name = os.path.splitext(os.path.basename(transcript_file_path))[0]
     episode_dir = chunks_dir
@@ -50,13 +55,13 @@ def check_timestamps_and_slice(audio_file_path, transcript_file_path, chunks_dir
     audio = AudioSegment.from_file(audio_file_path)
     for i, (start_ms, end_ms) in enumerate(timestamps):
         sliced_audio = audio[start_ms:end_ms]
-        print("sliced chunk")
         output_file_name = f'sliced_chunk_{i+1}.wav'
         output_file_path = os.path.join(episode_dir, output_file_name)
         sliced_audio.export(output_file_path, format="wav")
         timestamp_records.append(f"{output_file_name} {start_ms / 1000.0}")
+        logging.info(f"Sliced audio chunk saved as {output_file_name}")
 
-    print(f"Slicing complete for {episode_name}.")
+    logging.info(f"Slicing complete for {episode_name}.")
 
 def main():
     parser = argparse.ArgumentParser(description='Audio Processing with Whisper and PyDub')
@@ -73,43 +78,31 @@ def main():
     device = "cuda:0" if torch.cuda.is_available() else "cpu"
     torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
     model_id = "openai/whisper-base"
-    model = AutoModelForSpeechSeq2Seq.from_pretrained(
-        model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
-    )
+    model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True)
     model.to(device)
     processor = AutoProcessor.from_pretrained(model_id)
-    pipe = pipeline(
-        "automatic-speech-recognition",
-        model=model,
-        tokenizer=processor.tokenizer,
-        feature_extractor=processor.feature_extractor,
-        max_new_tokens=128,
-        chunk_length_s=30,
-        batch_size=16,
-        return_timestamps=True,
-        torch_dtype=torch_dtype,
-        device=device,
-    )
+    pipe = pipeline("automatic-speech-recognition", model=model, tokenizer=processor.tokenizer, feature_extractor=processor.feature_extractor, max_new_tokens=128, chunk_length_s=30, batch_size=16, return_timestamps=True, torch_dtype=torch_dtype, device=device)
 
     if os.path.isdir(args.path):
         audio_files = [f for f in os.listdir(args.path) if f.endswith(('.mp3', '.wav', '.m4a'))]
+        logging.info(f"Processing {len(audio_files)} audio files in directory {args.path}")
         for filename in tqdm(audio_files, desc="Processing audio files"):
             file_path = os.path.join(args.path, filename)
             transcript_file_path = transcribe_audio(file_path, pipe, transcripts_dir)
             if transcript_file_path:
                 check_timestamps_and_slice(file_path, transcript_file_path, chunks_dir, args.jump_threshold, timestamp_records)
     elif os.path.isfile(args.path):
+        logging.info(f"Processing single audio file at {args.path}")
         transcript_file_path = transcribe_audio(args.path, pipe, transcripts_dir)
         if transcript_file_path:
             check_timestamps_and_slice(args.path, transcript_file_path, chunks_dir, args.jump_threshold, timestamp_records)
     else:
-        print("The provided path does not exist.")
+        logging.error("The provided path does not exist.")
 
-    # Write timestamp records to a file
     with open("timestamps.txt", 'w') as f:
         for record in timestamp_records:
             f.write(record + '\n')
-    print("Timestamps recorded in 'timestamps.txt'.")
+    logging.info("Timestamps recorded in 'timestamps.txt'.")
 
 if __name__ == "__main__":
-    main()
+    main()