Update app.py

app.py

@@ -5,23 +5,24 @@ import yt_dlp as youtube_dl
 from transformers import pipeline
 from transformers.pipelines.audio_utils import ffmpeg_read
 from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
+from faster_whisper import WhisperModel
 
 
 import tempfile
 import os
 
-MODEL_NAME = "openai/whisper-medium"
+MODEL_NAME = "medium"
 BATCH_SIZE = 8
 FILE_LIMIT_MB = 1000
 
 device = 0 if torch.cuda.is_available() else "cpu"
 
-pipe = pipeline(
-    task="automatic-speech-recognition",
-    model=MODEL_NAME,
-    chunk_length_s=30,
-    device=device,
-)
+# pipe = pipeline(
+#     task="automatic-speech-recognition",
+#     model=MODEL_NAME,
+#     chunk_length_s=30,
+#     device=device,
+# )
 
 model = MBartForConditionalGeneration.from_pretrained("sanjitaa/mbart-many-to-many")
 tokenizer = MBart50TokenizerFast.from_pretrained("sanjitaa/mbart-many-to-many")
@@ -30,7 +31,9 @@ def translate(inputs, task):
     if inputs is None:
         raise gr.Error("No audio file submitted! Please upload or record an audio file before submitting your request.")
 
-    text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
+    #text = pipe(inputs, batch_size=BATCH_SIZE, generate_kwargs={"task": task}, return_timestamps=True)["text"]
+        ts_model = WhisperModel(MODEL_NAME, device = device, compute_type = "int8")
+        text = ts_model.transcribe(inputs, task = "translate")
 
     encoded_text = tokenizer(text, return_tensors="pt")
     tokenizer.src_lang = "en_XX"