@hht @isong Thank you for sharing your solution and issues. I'm also trying to run this "tune_relay_vta.py". After downloading tvm v0.7 released 2020-10-02, I've revised "vta_config.json" (TARGET: "sim" -> "pynq" from /3rdparty/vta-hw/config/vta_config.json) and saved a bitstream file "1x16_i8w8a32_15_15_18_17.bit" in the local directory to avoid the wrong link issue. Then, in the build directory, I've designated the llvm route of "config.cmake" for the host PC. Likewise, I've revised USE_VTA_FPGA OFF -> ON of "config.cmake" for the pynq board. There seemed to be no problems with the device connection through the tracker and the bitstream file connection.
However, when I run the "tune_relay_vta.py" (I've commented out the code @hht explained.), it doesn't increase GFLOPS (always 0.00/ 0.00 GFLOPS). Could you give me some advice? Do I need to change another code? --- [Visit Topic](https://discuss.tvm.apache.org/t/vta-workaround-for-autotuning-with-one-pynq-z1-board/8091/10) to respond. You are receiving this because you enabled mailing list mode. To unsubscribe from these emails, [click here](https://discuss.tvm.apache.org/email/unsubscribe/f9ae8e670b486463bf5e7674db33f481e39a40d137c780aed236ad77b431faa0).
