Update app.py

app.py

@@ -8,8 +8,8 @@ from transformers import (
 )
 
 # ================= CONFIG =================
-MODEL_ID = "Neon-AI/Kushina"   # your HF repo
-MAX_NEW_TOKENS = 1024          # generation cap (safe for CPU)
+MODEL_ID = "Neon-AI/Kushina"   # Hugging Face repo
+MAX_NEW_TOKENS = 16384          # high enough to avoid cutting essays
 TEMPERATURE = 0.7
 TOP_P = 0.9
 # ==========================================
@@ -27,7 +27,7 @@ def load_model():
     )
     model = AutoModelForCausalLM.from_pretrained(
         MODEL_ID,
-        torch_dtype=torch.float32,
+        torch_dtype=torch.float32,  # float32 on CPU
         device_map="cpu"
     )
     model.eval()