Cant convert custom onnx model to rknn

• quocan123

  Posts: 18

  Joined: 2025-03-12 2:38

  2025-04-17 8:22

  Hi, i'm try to convert my custom model to rknn for RV1106. But I got this error. How to fix it ?

  Code: Select all

  I rknn-toolkit2 version: 2.3.0
  --> Config model
  done
  --> Loading model
  I Loading : 100%|██████████████████████████████████████████████| 120/120 [00:00<00:00, 42897.51it/s]
  done
  --> Building model
  I OpFusing 0: 100%|██████████████████████████████████████████████| 100/100 [00:00<00:00, 531.79it/s]
  I OpFusing 1 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 277.26it/s]
  I OpFusing 0 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 177.67it/s]
  I OpFusing 1 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 169.86it/s]
  I OpFusing 2 : 100%|█████████████████████████████████████████████| 100/100 [00:00<00:00, 108.21it/s]
  W build: found outlier value, this may affect quantization accuracy
                          const name               abs_mean    abs_std     outlier value
                          model.0.conv.weight      0.69        0.87        -11.014
  I GraphPreparing : 100%|███████████████████████████████████████| 179/179 [00:00<00:00, 10823.93it/s]
  I Quantizating : 100%|████████████████████████████████████████████| 179/179 [00:05<00:00, 35.22it/s]
  W build: The default input dtype of 'images' is changed from 'float32' to 'int8' in rknn model for performance!
                         Please take care of this change when deploy rknn model with Runtime API!
  W build: The default output dtype of 'output0' is changed from 'float32' to 'int8' in rknn model for performance!
                        Please take care of this change when deploy rknn model with Runtime API!
  I rknn building ...
  E RKNN: [08:18:22.083] Squeezed bytes overflow!
  Aborted (core dumped)
  
  

  • Quote