Toybrick

标题: pt导出onnx并推理ok，但是onnx转rknn过程build报错_M_range_check [打印本页]
作者: htzhao    时间: 2023-9-22 17:47
标题: pt导出onnx并推理ok，但是onnx转rknn过程build报错_M_range_check
本帖最后由 htzhao 于 2023-9-22 17:49 编辑
  1. D RKNN: [17:35:15.277] >>>>>> start: N4rknn20RKNNMultiSurfacePassE
  2. D RKNN: [17:35:15.277] <<<<<<<< end: N4rknn20RKNNMultiSurfacePassE
  3. D RKNN: [17:35:15.277] >>>>>> start: N4rknn14RKNNTilingPassE
  4. E build: Catch exception when building RKNN model!
  5. E build: Traceback (most recent call last):
  6. E build:   File "rknn/api/rknn_base.py", line 2012, in rknn.api.rknn_base.RKNNBase.build
  7. E build:   File "rknn/api/rknn_base.py", line 384, in rknn.api.rknn_base.RKNNBase._generate_rknn
  8. E build:   File "rknn/api/rknn_base.py", line 348, in rknn.api.rknn_base.RKNNBase._build_rknn
  9. E build: IndexError: vector::_M_range_check: __n (which is 2) >= this->size() (which is 3)
复制代码
看了下日志，不知是否和这个有关


  1. I
  2. I Output[1140] shape with str value may cause error, replace ['Mul1140_dim_0', 'Mul1140_dim_1', 'Mul1140_dim_2', 'Mul1140_dim_3'] with [1, 1, 720, 1280].
  3. I Output[1142] shape with str value may cause error, replace ['Mul1142_dim_0', 'Mul1142_dim_1', 'Mul1142_dim_2', 'Mul1142_dim_3'] with [1, 1, 180, 320].
  4. I correct_ops ...
  5. I correct_ops done.
复制代码



欢迎光临 Toybrick (https://t.rock-chips.com/) Powered by Discuz! X3.3