Update app.py

app.py

@@ -249,12 +249,12 @@ model_location = "./MM_FT_C1"
 
 model = MultimodalPhiModel.from_pretrained(model_location).to(device)
 
-# model_name = "microsoft/Phi-3.5-mini-instruct"
-# base_phi_model = AutoModelForCausalLM.from_pretrained(
-#     model_name,
-#     torch_dtype=torch.bfloat16,
-#     trust_remote_code=True,
-# ).to(device)
+model_name = "microsoft/Phi-3.5-mini-instruct"
+base_phi_model = AutoModelForCausalLM.from_pretrained(
+    model_name,
+    torch_dtype=torch.bfloat16,
+    trust_remote_code=True,
+).to(device)
 
 def getStringAfter(output, start_str):
     if start_str in output:
@@ -308,7 +308,7 @@ def generateOutput(image_path, audio_path, context_text, question, max_length=3)
     # base_phi_model.generate(start_tokens, max_length=2, do_sample=False, pad_token_id=tokenizer.pad_token_id)
 
     output_text = tokenizer.decode(
-        model.base_phi_model.generate(start_tokens, attention_mask=attention_mask, max_length=1024, do_sample=False, pad_token_id=tokenizer.pad_token_id)[0],
+        base_phi_model.generate(start_tokens, attention_mask=attention_mask, max_length=1024, do_sample=False, pad_token_id=tokenizer.pad_token_id)[0],
         skip_special_tokens=True
     )