Update app.py

app.py

@@ -36,9 +36,10 @@ SAM_ENCODER_VERSION = "vit_h"
 SAM_CHECKPOINT_PATH = "./sam_vit_h_4b8939.pth"
 
 # Building GroundingDINO inference model
-groundingdino_model  = load_model(model_config_path=GROUNDING_DINO_CONFIG_PATH, model_checkpoint_path=GROUNDING_DINO_CHECKPOINT_PATH, device="cpu")
+groundingdino_model  = load_model(model_config_path=GROUNDING_DINO_CONFIG_PATH, model_checkpoint_path=GROUNDING_DINO_CHECKPOINT_PATH, device="cuda")
 # Building SAM Model and SAM Predictor
 sam = build_sam(checkpoint=SAM_CHECKPOINT_PATH)
+sam.to(device="cuda")
 sam_predictor = SamPredictor(sam)
 
 def transform_image(image_pil):
@@ -128,7 +129,7 @@ def get_mask(image, label):
     sam_predictor.set_image(image)
 
     transformed_boxes = sam_predictor.transform.apply_boxes_torch(
-        boxes_filt, image.shape[:2])
+        boxes_filt, image.shape[:2]).to("cuda")
 
     masks, _, _ = sam_predictor.predict_torch(
         point_coords=None,
@@ -359,7 +360,7 @@ with gr.Blocks() as demo:
                 text_prompt = gr.Textbox(label="Label")
 
         with gr.Column(scale=1):
-            baseline_gallery = gr.Gallery(label='Output', show_label=True, elem_id="gallery", height=721, columns=1)
+            baseline_gallery = gr.Gallery(label='Output', show_label=True, elem_id="gallery", height=800, columns=1)
             with gr.Accordion("Advanced Option", open=True):
                 seed = gr.Slider(label="Seed", minimum=-1, maximum=999999999, step=1, value=666)
                 gr.Markdown("### Guidelines")