Update app.py

app.py

@@ -43,13 +43,13 @@ phi_model_name = "microsoft/phi-2"
 text_tokenizer = AutoTokenizer.from_pretrained(
     phi_model_name, trust_remote_code=True)
 with torch.no_grad():
-    tuned_phi2 = AutoModelForCausalLM.from_pretrained(
-        "stage2_adaptor", trust_remote_code=True,
-    )
     base_phi2_text = AutoModelForCausalLM.from_pretrained(
         phi_model_name, trust_remote_code=True,
         device_map="auto", torch_dtype=torch.float16
     )
+    tuned_phi2 = AutoModelForCausalLM.from_pretrained(
+        "stage2_adaptor", trust_remote_code=True,
+    ).to("cpu")
 print("phi2 model loaded")
 
 audio_model_name = "openai/whisper-small"