Upload /app.py with huggingface_hub

app.py

@@ -185,7 +185,6 @@ def get_visual_prompts_embeddings(
         boxes_prompts[:, [0, 2]] = boxes_prompts[:, [0, 2]] / max(height, width) * 1024
         boxes_prompts[:, [1, 3]] = boxes_prompts[:, [1, 3]] / max(height, width) * 1024
         boxes_prompts = torch.from_numpy(boxes_prompts)
-        boxes_prompts = torch.from_numpy(boxes_prompts)
         boxes_prompts = boxes_prompts.cuda()
         # using <region> token
         region_token_id = omg_llava.region_token_idx
@@ -193,7 +192,7 @@ def get_visual_prompts_embeddings(
         boxes_mark_embedding = get_points_embeddings(
             boxes_prompts, input_ids,
             1024, 1024,
-            region_token_id)
+            region_token_id, mode='point')
     return points_mark_embedding, boxes_mark_embedding
 
 def inference(input_str, all_inputs, follow_up):