Update app.py

app.py

@@ -27,6 +27,7 @@ def transcribe_files(audio_files):
     outdir = os.path.join(workdir, "transcripts")
     os.makedirs(outdir, exist_ok=True)
     
+    # Create pipeline
     pipe = pipeline(
         "automatic-speech-recognition",
         model="palli23/whisper-small-sam_spjall",
@@ -34,6 +35,9 @@ def transcribe_files(audio_files):
         device=0,
     )
     
+    # Force Icelandic language using tokenizer
+    forced_decoder_ids = pipe.tokenizer.get_decoder_prompt_ids(language="is", task="transcribe")
+    
     for file in audio_files:
         audio_path = file.name
         base = os.path.splitext(os.path.basename(audio_path))[0]
@@ -45,8 +49,7 @@ def transcribe_files(audio_files):
             batch_size=8,
             return_timestamps=False,
             generate_kwargs={
-                "language": "is",
-                "task": "transcribe",
+                "forced_decoder_ids": forced_decoder_ids,
                 "num_beams": 5,
                 "repetition_penalty": 1.2,
                 "no_repeat_ngram_size": 3,