Update app.py

app.py

@@ -60,17 +60,20 @@ async def transcribe_chunked_audio(audio_file: UploadFile, task: str = "transcri
         raise HTTPException(status_code=400, detail=f"File size exceeds file size limit. Got file of size {file_size_mb:.2f}MB for a limit of {FILE_LIMIT_MB}MB.")
 
     try:
+        logger.debug(f"Opening audio file: {audio_file.filename}")
         with open(audio_file.filename, "rb") as f:
             inputs = f.read()
     except Exception as e:
         logger.error("Error reading audio file:", exc_info=True)
         raise HTTPException(status_code=500, detail="Error reading audio file")
 
+    logger.debug("Performing ffmpeg read on audio file")
     inputs = ffmpeg_read(inputs, pipeline.feature_extractor.sampling_rate)
     inputs = {"array": inputs, "sampling_rate": pipeline.feature_extractor.sampling_rate}
     logger.debug("Done loading audio file")
 
     try:
+        logger.debug("Calling tqdm_generate to transcribe audio")
         text, runtime = tqdm_generate(inputs, task=task, return_timestamps=return_timestamps)
     except Exception as e:
         logger.error("Error transcribing audio:", exc_info=True)
@@ -97,6 +100,7 @@ async def transcribe_youtube(yt_url: str = Form(...), task: str = "transcribe",
             raise HTTPException(status_code=500, detail="Error downloading YouTube audio")
 
         try:
+            logger.debug(f"Opening downloaded audio file: {filepath}")
             with open(filepath, "rb") as f:
                 inputs = f.read()
         except Exception as e:
@@ -108,6 +112,7 @@ async def transcribe_youtube(yt_url: str = Form(...), task: str = "transcribe",
     logger.debug("Done loading YouTube file")
 
     try:
+        logger.debug("Calling tqdm_generate to transcribe YouTube audio")
         text, runtime = tqdm_generate(inputs, task=task, return_timestamps=return_timestamps)
     except Exception as e:
         logger.error("Error transcribing YouTube audio:", exc_info=True)
@@ -121,17 +126,19 @@ def tqdm_generate(inputs: dict, task: str, return_timestamps: bool):
     num_samples = len(all_chunk_start_idx)
     num_batches = math.ceil(num_samples / BATCH_SIZE)
 
+    logger.debug("Preprocessing audio for inference")
     dataloader = pipeline.preprocess_batch(inputs, chunk_length_s=CHUNK_LENGTH_S, batch_size=BATCH_SIZE)
     model_outputs = []
     start_time = time.time()
     logger.debug("Transcribing...")
     # iterate over our chunked audio samples - always predict timestamps to reduce hallucinations
     for batch in dataloader:
+        logger.debug(f"Processing batch of {len(batch)} samples")
         model_outputs.append(pipeline.forward(batch, batch_size=BATCH_SIZE, task=task, return_timestamps=True))
     runtime = time.time() - start_time
     logger.debug("Done transcription")
 
-    logger.debug("Post-processing...")
+    logger.debug("Post-processing transcription results")
     try:
         post_processed = pipeline.postprocess(model_outputs, return_timestamps=True)
     except Exception as e: