Ultra-low power and high-performance AI processor GAP8 › Forums › GAP8 developers’ group › SDK (Software Development Kit) › Using my own mnist trained weights with mbed os example
- Joris QuerryParticipantJanuary 10, 2019 at 3:56 pmPost count: 7
I’v been training the mnist example from ~/gap_sdk/tf2gap8/examples/mnist
The result is a bunch of files into ~/gap_sdk/tf2gap8/examples/mnist/data
Then I used the ~/gap_sdk/tf2gap8/tf2gap8/tf2gap8.py script which results in another bunch of files into : ~gap_sdk/tf2gap8/examples/mnist/data/GAP8Code
Now, I would like to use these files within the gapuino, like in the ~/gap_sdk/examples/mbed-examples/test_autotiler/Mnist/MnistCoeffs.def example.
But I don’t find a direct way to convert/edit/copy my files.
Do you have any tutorial on this point ?
- Joel CamboniParticipantJanuary 14, 2019 at 3:59 pmPost count: 3
The tf2gap8.py script is launched by the make file.
After training (that generates the files in the ./data directory), you have to do “make clean all run” in the examples/mnist directory.
The weights and bias will be generated in the GAP8Code/weight_bias.c file.
The generated code will be compiled and executed on the Gapuino board.
We will post a more detailed documentation on the flow very soon.
please keep me posted of your progress.
- Joris QuerryParticipantMarch 21, 2019 at 4:04 pmPost count: 7
I just found this documentation readme :
You should had a link to it on the tf2gap8 github page ! 🙂
- You must be logged in to reply to this topic.