Update app.py

app.py

@@ -28,8 +28,8 @@ class _MLPVectorProjector(nn.Module):
 model_name = "microsoft/phi-2"
 
 with torch.no_grad():
-    phi2_text = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True, device_map="auto")
-    
+    phi2_text = AutoModelForCausalLM.from_pretrained(model_name, trust_remote_code=True, device_map="auto",dtype=torch.float16)
+     
 tokenizer_text = AutoTokenizer.from_pretrained(model_name, trust_remote_code=True)
 
 ## Audio model
@@ -86,9 +86,10 @@ def example_inference(input_text, count, image, img_qn, audio):
 
 def textMode(text, count):
     count = int(count)
+    text = "Question: " + text + "Answer: "
     inputs = tokenizer_text(text, return_tensors="pt", return_attention_mask=False)
     prediction = tokenizer_text.batch_decode(
-    phi2_text.generate(
+    phi2_finetuned.generate(
         **inputs, 
         max_new_tokens=count,
         bos_token_id=tokenizer_text.bos_token_id,