adding zerogpu

app.py

@@ -37,18 +37,23 @@ if model_provider.lower() == "openai":
     )
 else:
     MODEL_NAME = "meta-llama/Llama-3.3-70B-Instruct"
-    # Initialize Hugging Face InferenceClient
+    # Initialize Hugging Face InferenceClient with GPU support
     hf_client = InferenceClient(
         model=MODEL_NAME,
-        api_key=os.environ.get("HF_TOKEN")
+        api_key=os.environ.get("HF_TOKEN"),
+        timeout=120  # Increased timeout for GPU inference
     )
 
 # Load the Hugging Face dataset
 dataset = load_dataset('tosin2013/autogen', streaming=True)
 dataset = Dataset.from_list(list(dataset['train']))
 
-# Initialize embeddings
-embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-MiniLM-L6-v2")
+# Initialize embeddings with GPU support if available
+device = "cuda" if torch.cuda.is_available() else "cpu"
+embeddings = HuggingFaceEmbeddings(
+    model_name="sentence-transformers/all-MiniLM-L6-v2",
+    model_kwargs={"device": device}
+)
 
 # Extract texts from the dataset
 texts = dataset['input']