Python api simulator

Eric

I had a yolo base model (Pytorch), already converted by onnx2onnx2.py.

In this code , it can get good result

km = ktc.ModelConfig(19, "0001", "520", onnx_path="/data1/mbv2_yolo_output.onnx")

But I change it to 720

km = ktc.ModelConfig(19, "0001", "720", onnx_path="/data1/mbv2_yolo_output.onnx")

I got this message

Fix point analysis done. Save bie model to '/data1/output.bie'
terminate called after throwing an instance of 'nonstd::variants::bad_variant_access'
 what(): bad variant access
Aborted (core dumped)

And how can I solve this problem ?

total flow code :

km = ktc.ModelConfig(19, "0001", "720", onnx_path="/data1/mbv2_yolo_output.onnx")
eval_result = km.evaluate()
print("\nNpu performance evaluation result:\n" + str(eval_result))

bie_model_path = km.analysis({"input.1": img_list})
print("\nFix point analysis done. Save bie model to '" + str(bie_model_path) + "'")
# check nef radix from quantization data
radix = ktc.get_radix(img_list)
print(radix,bie_model_path)
# bie inference 
out_data = ktc.kneron_inference([in_data], bie_file=bie_model_path, input_names=["input.1"], radix=radix,platform=720)

# bie output data processing
det_res = postprocess(out_data, [input_image.size[1], input_image.size[0]])
draw_bboxes(input_image,det_res)
im = input_image.save("/home/data/onnx_result.jpg")

Env :

kneron/toolchain:latest

ONNX file

https://github.com/eric612/nef_model_zoo/blob/main/mbv2_yolo_voc/mbv2_yolo_output.onnx

anchors :

anchors.txt

Eric

Solved , please delete this post