IMX8MP EVK / efficientdet-lite0 NPU execution fail

Kei-Ueda — Thu, 28 Apr 2022 08:23:56 GMT

I'm trying to use efficientdet-lite0(TensorFlow Hub (tfhub.dev)) on IMX8MP / NPU but currently I'm facing the following Ovxlib error (CPU execution is OK).

========

E [_init_tensor:408]Create vx tensor fail.
E [_create_tensor:478]Create vx tensor fail.
E [vsi_nn_CreateTensorFromData:1197]Create tensor from data fail.
benchmark_model: ovxlib_delegate.cpp:809: int nnrt::OvxlibDelegate::addTensor(vsi_nn_graph_t*, vsi_nn_tensor_attr_t*, size_t, const void*, bool): Assertion `false' failed.
Aborted

========

AlthoughI know efficientdet-lite0 is not included in Supported test model (i.MX Machine Learning User's Guide (nxp.com) ), there is no special operator in efficientdet-lite0. so I expected NPU execution without any errors.

I'd like to grasp the countermeasure of this Ovxlib error.

topic IMX8MP EVK / efficientdet-lite0 NPU execution fail in i.MX Processors

IMX8MP EVK / efficientdet-lite0 NPU execution fail