RuntimeError: ArrayRef: invalid index Index = 15; Length = 15

[zxcvbml]

[03/25 09:58:22 detectron2]: load Model: ../../../data/DCIS/exp_pth/FCOS_R_50_1x.pth Traceback (most recent call last): File "onnx/export_model_to_onnx.py", line 302, in main() File "onnx/export_model_to_onnx.py", line 294, in main output_names=output_names, File "/root/miniconda3/envs/detectron2/lib/python3.7/site-packages/torch/onnx/init.py", lin e 276, in export custom_opsets, enable_onnx_checker, use_external_data_format) File "/root/miniconda3/envs/detectron2/lib/python3.7/site-packages/torch/onnx/utils.py", line 9 4, in export use_external_data_format=use_external_data_format) File "/root/miniconda3/envs/detectron2/lib/python3.7/site-packages/torch/onnx/utils.py", line 7 01, in _export dynamic_axes=dynamic_axes) File "/root/miniconda3/envs/detectron2/lib/python3.7/site-packages/torch/onnx/utils.py", line 4 88, in _model_to_graph torch._C._jit_pass_onnx_assign_output_shape(graph, output_tensors, out_desc, _onnx_shape_infe rence)RuntimeError: ArrayRef: invalid index Index = 15; Length = 15

torch 1.8.1+cu101 python 3.7.4

the BlendMask config as follow： BASE: "Base-550.yaml" INPUT: MIN_SIZE_TRAIN: (256, 288, 320, 352, 384, 416, 448, 480, 512, 544, 576, 608) MAX_SIZE_TRAIN: 900 MAX_SIZE_TEST: 736 MIN_SIZE_TEST: 512 MODEL: WEIGHTS: "detectron2://ImageNetPretrained/MSRA/R-50.pkl" RESNETS: DEPTH: 50 NORM: "SyncBN" BACKBONE: FREEZE_AT: -1 FCOS: NORM: "SyncBN" SOLVER: STEPS: (300000, 340000) MAX_ITER: 360000 OUTPUT_DIR: "output/blendmask/RT_R_50_4x-bn-head"

----------------------------------------------------------------------------------------------------------------

I have tried many pth model including FCOS, BlensMasks , that Will make the same ArrayRef error. I hope to get some suggestions. Thank you a lot.

[JJC6050]

I also encountered the same problem. Have you solved it? RuntimeError: ArrayRef: invalid index Index = 21; Length = 21

[zxcvbml]

I also encountered the same problem. Have you solved it? RuntimeError: ArrayRef: invalid index Index = 21; Length = 21

首先查看模型的FCOS head的cls_tower和bbox_tower是否为空，我的某些模型会出现为“ (cls_tower): Sequential()，(bbox_tower): Sequential()”，然后选用了不为空的权重模型， 发现无法进入如下if语句 if yield_bbox_towers: bbox_towers.append(bbox_tower) 去掉if 直接append。成功。（在/export_model_to_onnx.py下修改）

[JJC6050]

https://github.com/aim-uofa/AdelaiDet/issues/446