NVIDIA-AI-IOT / cuDLA-samples

YOLOv5 on Orin DLA
Other
167 stars 17 forks source link

Size of output tensor in Run function #33

Open mayulin0206 opened 3 months ago

mayulin0206 commented 3 months ago

matx::tensor_t<matx::matxFp16, 5> mHead1Input1 = matx::make_tensor((matx::matxFp16)src[0], {1, 16, 84, 84, 16}); matx::tensor_t<matx::matxFp16, 5> mHead1Output1 = matx::make_tensor((matx::matxFp16)mTemp, {1, 16, 16, 84, 84});

The output shape is {1, 16, 84, 84, 16} why the shape is {1, 16, 84, 84, 16}? other shape is okey? like{1, 8, 168,84, 16}