| 16 | from vision_reasoner.models.qwen_vl_cot import QwenVLCoTModel |
| 17 | |
| 18 | def parse_args(): |
| 19 | parser = argparse.ArgumentParser() |
| 20 | parser.add_argument("--model", type=str, default="vision_reasoner") |
| 21 | parser.add_argument("--model_path", type=str, default="pretrained_models/VisionReasoner-7B") |
| 22 | parser.add_argument("--task_router_model_path", type=str, default="pretrained_models/TaskRouter-1.5B") |
| 23 | parser.add_argument("--segmentation_model_path", type=str, default="facebook/sam2-hiera-large") |
| 24 | parser.add_argument("--output_path", type=str, required=True) |
| 25 | parser.add_argument("--test_data_path", type=str, required=True) |
| 26 | parser.add_argument("--batch_size", type=int, default=1) |
| 27 | |
| 28 | # for parallel evaluation |
| 29 | parser.add_argument("--idx", type=int, required=True) |
| 30 | parser.add_argument("--num_parts", type=int, required=True) |
| 31 | return parser.parse_args() |
| 32 | |
| 33 | def compute_bbox_iou(bboxes1, bboxes2): |
| 34 | """ |