Update app.py

app.py

@@ -23,7 +23,7 @@ MODEL = 'llama3-groq-70b-8192-tool-use-preview'
 
 # Load MiniCPM-V-2_6 with 4-bit quantization
 text_model = AutoModel.from_pretrained('openbmb/MiniCPM-V-2', trust_remote_code=True,
-                                       device_map="auto", dtype=torch.float16)
+                                       device_map="auto", torch_dtype=torch.bfloat16)
 tokenizer = AutoTokenizer.from_pretrained('openbmb/MiniCPM-V-2', trust_remote_code=True)
 
 tts_model = ParlerTTSForConditionalGeneration.from_pretrained("parler-tts/parler-tts-large-v1").to('cuda')
@@ -176,6 +176,7 @@ def initialize_tools():
 @spaces.GPU()
 # Gradio Interface
 def main_interface(user_prompt, image=None, video=None, audio=None, doc=None, voice_only=False):
+    text_model = text_model.to(device='cuda', dtype=torch.bfloat16)
     response = handle_input(user_prompt, image=image, video=video, audio=audio, doc=doc)
     if voice_only:
         audio_file = play_voice_output(response)