Add HF_TOKEN support for gated model access

app.py

@@ -8,23 +8,27 @@ import os
 app = Flask(__name__)
 CORS(app)
 
+# Hugging Face token'ı environment variable'dan al
+HF_TOKEN = os.environ.get("HF_TOKEN")
+
 # Model adları
 MODEL_NAME = "ilkeileri/gemma-sales-comprehensive"
 BASE_MODEL = "google/gemma-1.1-2b-it"
 
 print("Loading tokenizer...")
-tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True)
+tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL, trust_remote_code=True, token=HF_TOKEN)
 
 print("Loading base model...")
 base_model = AutoModelForCausalLM.from_pretrained(
     BASE_MODEL,
     dtype=torch.float16,
     low_cpu_mem_usage=True,
-    trust_remote_code=True
+    trust_remote_code=True,
+    token=HF_TOKEN
 )
 
 print("Loading LoRA adapters...")
-model = PeftModel.from_pretrained(base_model, MODEL_NAME)
+model = PeftModel.from_pretrained(base_model, MODEL_NAME, token=HF_TOKEN)
 model.eval()
 
 print("Model loaded successfully!")