diff --git a/biotrack/tracker.py b/biotrack/tracker.py
index 3cfdada..ae92342 100644
--- a/biotrack/tracker.py
+++ b/biotrack/tracker.py
@@ -17,9 +17,6 @@
 from biotrack.track import Track
 from biotrack.logger import create_logger_file, info, debug
 
-DEFAULT_DEVICE = "cuda" if torch.cuda.is_available() else "mps" if torch.backends.mps.is_available() else "cpu"
-
-
 class BioTracker:
     def __init__(self, image_width: int, image_height: int, device_id: int = 0, **kwargs):
         self.logger = create_logger_file()
@@ -36,13 +33,13 @@ def __init__(self, image_width: int, image_height: int, device_id: int = 0, **kw
         self.track_model = CoTrackerPredictor(checkpoint=None)
         self.track_model.model = model.model
         self.track_model.step = model.model.window_len // 2
-
-        self.device = torch.device(f"cuda:{device_id}" if torch.cuda.is_available() else "cpu")
+        self.device_id = kwargs.get("gpu_id", 0)
+        self.device = torch.device(f"cuda:{self.device_id}" if torch.cuda.is_available() else "cpu")
         self.track_model.model.to(self.device)
 
         # Initialize the model for computing crop embeddings
         model_name = kwargs.get("vits_model", ViTWrapper.DEFAULT_MODEL_NAME)
-        self.vit_wrapper = ViTWrapper(DEFAULT_DEVICE, device_id=device_id, model_name=model_name)
+        self.vit_wrapper = ViTWrapper(device_id=device_id, model_name=model_name)
 
     def update_trackers_queries(self, frame_num: int, keypoints: np.array, labels: List[str], scores: np.array, coverages: np.array, boxes: np.array,  d_emb: np.array, **kwargs):
         """