default_concurrency_limit

app.py

@@ -6,11 +6,25 @@ from faster_whisper import WhisperModel
 from languages import get_language_names, get_language_from_name
 from subtitle_manager import Subtitle
 from pathlib import Path
+import psutil
+import torch
 
 logging.basicConfig(level=logging.INFO)
 last_model = None
 model = None
 
+
+def get_workers_count():
+    if torch.cuda.is_available():
+        memory = torch.cuda.get_device_properties(0).total_memory
+    else:
+        memory = psutil.virtual_memory().total
+    
+    workers = int(memory / 2_000_000_000)
+    logging.info(f"workers:{workers}")
+    logging.info(f"memory:{memory/ 1_000_000_000} GB")
+    return workers
+    
 def transcribe_webui_simple_progress(modelName, languageName, urlData, multipleFiles, microphoneData, task,
                                     chunk_length, compute_type, beam_size, vad_filter, min_silence_duration_ms,
                                     progress=gr.Progress()):
@@ -132,5 +146,5 @@ with gr.Blocks(title="Fast Whisper WebUI") as demo:
     )
 
 if __name__ == "__main__":
-    demo.launch(share=True)
+    demo.launch(share=True, default_concurrency_limit=get_workers_count())