Update app.py

app.py

@@ -3,7 +3,6 @@ import torch
 from fastapi import FastAPI, Request
 from transformers import AutoTokenizer, AutoModelForCausalLM
 from peft import PeftModel
-from huggingface_hub import login
 from typing import Dict, Any
 
 # Hugging Face token
@@ -11,8 +10,9 @@ HF_TOKEN = os.getenv("HF_TOKEN")
 if not HF_TOKEN:
     raise ValueError("HF_TOKEN environment variable not set!")
 
-# Login to Hugging Face Hub
-login(token=HF_TOKEN)
+# Cache dizinini ayarla (yazma izni olan bir dizin)
+os.environ["HF_HOME"] = "/tmp/huggingface"
+os.environ["TRANSFORMERS_CACHE"] = "/tmp/huggingface"
 
 # Model IDs
 BASE_MODEL_ID = "google/gemma-1.1-2b-it"
@@ -21,10 +21,11 @@ LORA_MODEL_ID = "programci48/heytak-lora-v1"
 # Load models with error handling and optimizations
 def load_models() -> Dict[str, Any]:
     try:
-        # Load tokenizer
+        # Load tokenizer (login işlemi olmadan doğrudan token kullanarak)
         tokenizer = AutoTokenizer.from_pretrained(
             BASE_MODEL_ID,
-            token=HF_TOKEN
+            token=HF_TOKEN,
+            cache_dir="/tmp/huggingface"
         )
 
         # Load base model with memory optimization
@@ -34,7 +35,7 @@ def load_models() -> Dict[str, Any]:
             device_map="auto",
             token=HF_TOKEN,
             low_cpu_mem_usage=True,
-            offload_folder="offload"  # For CPU offloading if needed
+            cache_dir="/tmp/huggingface"
         )
 
         # Load LoRA adapter