change whisper version

README.md

@@ -1,6 +1,6 @@
 ---
 title: Parlons Nous
-python_version: 3.9.0
+python_version: 3.10
 emoji: 🏢
 colorFrom: purple
 colorTo: indigo

app.py

@@ -5,15 +5,30 @@ import numpy as np
 import torch
 from groq import Groq
 from transformers import pipeline
-from transformers.utils import is_flash_attn_2_available
 from TTS.api import TTS
 
-transcriber = pipeline("automatic-speech-recognition",
-                       model="openai/whisper-large-v3",
-                       torch_dtype=torch.float16,
-                       device="cuda:0",
-                       model_kwargs={"attn_implementation": "flash_attention_2"} if is_flash_attn_2_available() else {"attn_implementation": "sdpa"},
-                      )
+MODEL_NAME = "openai/whisper-large-v3"
+BATCH_SIZE = 8
+FILE_LIMIT_MB = 1000
+YT_LENGTH_LIMIT_S = 3600  # limit to 1 hour YouTube files
+
+device = 0 if torch.cuda.is_available() else "cpu"
+
+pipe = pipeline(
+    task="automatic-speech-recognition",
+    model=MODEL_NAME,
+    chunk_length_s=30,
+    device=device,
+)
+
+
+def use_pipe(inputs):
+    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": "transcribe"}, return_timestamps=True)["text"]
+    return  text
+
 
 groq_client = Groq(api_key=os.getenv('GROQ_API_KEY'))
 
@@ -36,7 +51,7 @@ def transcribe(stream, new_chunk):
         stream = np.concatenate([stream, y])
     else:
         stream = y
-    return stream, transcriber({"sampling_rate": sr, "raw": stream})["text"]
+    return stream, use_pipe(stream)
 
 def autocomplete(text):
     """

requirements.txt

@@ -1,6 +1,7 @@
 gradio
 groq
 numpy
+torch
 torchaudio
 transformers
 tts