Update app.py

app.py

@@ -37,7 +37,7 @@ clip_model = CLIPVisionModel.from_pretrained(clip_model_name).to(device)
 projection = torch.nn.Linear(clip_embed, phi_embed).to(device)
 resblock = SimpleResBlock(phi_embed).to(device)
 phi_model = AutoModelForCausalLM.from_pretrained(phi_model_name,trust_remote_code=True).to(device)
-audio_model = whisperx.load_model("large-v2", device, compute_type=compute_type)
+audio_model = whisperx.load_model("tiny", device, compute_type=compute_type)
 
 # load weights
 model_to_merge = PeftModel.from_pretrained(phi_model,'./model_chkpt/lora_adaptor')