Update app.py

app.py

@@ -1,6 +1,5 @@
 import os
 import torch
-import gradio as gr
 from transformers import AutoTokenizer, AutoModelForCausalLM, AutoConfig
 
 MODEL_ID = "Amey9766/llama32B-hospitality-review-triage"
@@ -8,15 +7,20 @@ HF_TOKEN = os.getenv("HF_TOKEN")
 
 tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, token=HF_TOKEN)
 
-# Load config and REMOVE any quantization config
 config = AutoConfig.from_pretrained(MODEL_ID, token=HF_TOKEN)
+
+# IMPORTANT: remove quantization_config cleanly
 if hasattr(config, "quantization_config"):
-    config.quantization_config = None
+    try:
+        delattr(config, "quantization_config")   # best option
+    except Exception:
+        config.quantization_config = {}          # fallback
 
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
     token=HF_TOKEN,
-    config=config,                         # <-- important
+    config=config,
+    quantization_config=None,                    # prevents auto-quant logic
     device_map="auto",
     dtype=torch.float16 if torch.cuda.is_available() else torch.float32
 )