Update app.py

app.py

@@ -4,35 +4,25 @@ from fastapi import FastAPI, Request
 from transformers import AutoTokenizer, AutoModelForCausalLM
 from peft import PeftModel
 
-# 🔐 HF_TOKEN değişkenini oku
-hf_token = os.getenv("HF_TOKEN")
+hf_token = os.getenv("HF_TOKEN")  # 🔑 Token ortam değişkeninden
 
-# 🔧 Model bilgisi
 base_model_id = "google/gemma-1.1-2b-it"
 lora_model_id = "programci48/heytak-lora-v1"
 
-# 📦 Model yükleme
+# 🔧 Tokenizer ve model yükleme
 tokenizer = AutoTokenizer.from_pretrained(base_model_id, token=hf_token)
-base_model = AutoModelForCausalLM.from_pretrained(
-    base_model_id,
-    torch_dtype=torch.float32,
-    device_map=None,
-    token=hf_token
-)
+base_model = AutoModelForCausalLM.from_pretrained(base_model_id, torch_dtype=torch.float32, device_map=None, token=hf_token)
 model = PeftModel.from_pretrained(base_model, lora_model_id, token=hf_token)
 model.eval()
 
-# 🚀 FastAPI app
 app = FastAPI()
 
 @app.post("/run/predict")
 async def predict(request: Request):
     data = await request.json()
     prompt = data["data"][0]
-
     inputs = tokenizer(prompt, return_tensors="pt")
     with torch.no_grad():
         outputs = model.generate(**inputs, max_new_tokens=100)
     response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-
     return {"data": [response]}