Yolov8 to rknn model

[Jay12138-max]

PC: x86 Ubuntu 22.04
rknn-toolkit2 version: 2.0.0b0+9bab5682

When I tried to convert my model "yolov8n.torchscript" or "yolov8.onnx" to "*.rknn", the following issues will occur.
I also tried this method , but not worked yet.
When I change target_platform from 'rk3588' to 'rk3568', there is no issue. How to solve this issue? Help!

---

I PtParse: 100%|█████████████████████████████████████████████████| 824/824 [00:06<00:00, 117.77it/s]
I Loading : 100%|████████████████████████████████████████████████| 149/149 [00:00<00:00, 395.82it/s]
I GraphPreparing : 100%|█████████████████████████████████████████| 177/177 [00:00<00:00, 672.25it/s]
I Quantizating : 100%|███████████████████████████████████████████| 177/177 [00:00<00:00, 421.73it/s]
W build: The default input dtype of 'x.3' 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 '1313' 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: [09:17:20.658] REGTASK: The bit width of field value exceeds the limit, target: v2, offset: 0x500c, shift = 0, limit: 0x1fff, value: 0x20cf
E RKNN: [09:17:20.658] REGTASK: The bit width of field value exceeds the limit, target: v2, offset: 0x500c, shift = 0, limit: 0x1fff, value: 0x20cf
E RKNN: [09:17:20.659] REGTASK: The bit width of field value exceeds the limit, target: v2, offset: 0x500c, shift = 0, limit: 0x1fff, value: 0x20cf
E RKNN: [09:17:20.705] failed to config argb mode layer!
已放弃 (核心已转储)

---

My conversation code is shown as follows:
(.onnx or .torchscript to .rknn)

from rknn.api import RKNN
import numpy

if __name__ == '__main__':
    rknn = RKNN()
    rknn.config(mean_values=[[0,0,0]],std_values=[[255,255,255]],target_platform='rk3588')

    # ret = rknn.load_onnx(model="best_22.onnx")
    ret  = rknn.load_pytorch(model = "best_22_rknnopt.torchscript",input_size_list=[[1,3,640,640]])
    rknn.build(do_quantization=True,dataset="dataset11.txt")
    print(ret)
    rknn.export_rknn('best33.rknn')
    rknn.release()

(.pt to .torchscript)

from ultralytics import YOLO 
model = YOLO("best_22.pt") 
model.export(format='RKNN')