• State Resolved
  • Locked Locked
  • Replies 1 reply
  • Subscribers 97 subscribers
  • Views 489 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

TI Deep Learing Library User-guide model imported by TIDL failed

clare zeng
Prodigy 60 points

We tried to import models (MobileNet v2, and Rsenet18 v2)  presented in $psdk/tidl_j7_01_00_00_00/ti_dl/docs/user_guide_html/md_tidl_models_info.html. All configuration files used the default. Then we validated the TIDL model in PC Emulation on a subset of ImgeNet, the results are as follows.  The TIDL models imported from Caffe works well, but the TIDL models imported from Onnx cannot classify images. 

  • 0
    TI__Mastermind 22145 points

    Mobile net v2 models trained on Pytorch and Tensorflow needs 16-bit for the current version of TIDL.

    We are working on improving our post-training quantization algorithm that would give better results with 8-bits also.

    Initial release this feature would be available by June mid.