feat: reduce load while model is first processing the data

app.py

@@ -6,11 +6,7 @@ import nltk
 
 @asynccontextmanager
 async def lifespan(app: FastAPI):
-    # Ensure punkt is available
-    nltk.download("punkt")
 
-    nltk.download('punkt_tab')
-    
     # Your model warmup
     warmup()
     yield

features/text_classifier/model_loader.py

@@ -5,7 +5,7 @@ from transformers import GPT2LMHeadModel, GPT2TokenizerFast, GPT2Config
 from huggingface_hub import snapshot_download
 import torch
 from dotenv import load_dotenv
-
+import nltk
 load_dotenv()
 REPO_ID = "Pujan-Dev/AI-Text-Detector"
 MODEL_DIR = "./models"
@@ -18,6 +18,11 @@ _model, _tokenizer = None, None
 
 def warmup():
     global _model, _tokenizer
+    # Ensure punkt is available
+    nltk.download("punkt")
+
+    nltk.download('punkt_tab')
+
     download_model_repo()
     _model, _tokenizer = load_model()
     logging.info("Its ready")