Added app.py and requirements.txt

app.py

@@ -19,11 +19,12 @@ print("Using device:", device)
 print("Loading base model...")
 base_model = AutoModelForCausalLM.from_pretrained(
     BASE_MODEL,
-    device_map=None,
+    device_map="auto",
+    load_in_8bit=True,
+    torch_dtype=torch.float16,
     low_cpu_mem_usage=True,
-    torch_dtype=torch.float16 if device == "cuda" else torch.float32,
-    token=HF_TOKEN
-).to(device)
+    use_auth_token=HF_TOKEN
+)
 
 print("Loading tokenizer...")
 tokenizer = AutoTokenizer.from_pretrained(