Executing run.py for TVM UMA Vanilla example fails

Following Vanilla tutorial and executing run.py results in the following error:

One or more operators have not been tuned. Please tune your model for better performance. Use DEBUG logging level to see more details. Generated files are in /tmp/tvm-debug-mode-tempdirs/2023-09-06T12-27-48___8bfi9oq5/00000 conv2d is not optimized for this platform. Traceback (most recent call last): File “vanilla_accelerator/run.py”, line 82, in main() File “vanilla_accelerator/run.py”, line 77, in main test_dir=str(export_directory), File “/home/hex/.local/lib/python3.7/site-packages/tvm-0.14.dev156+g909c8fab8-py3.7-linux-x86_64.egg/tvm/testing/aot.py”, line 906, in compile_and_run checker=checker, File “/home/hex/.local/lib/python3.7/site-packages/tvm-0.14.dev156+g909c8fab8-py3.7-linux-x86_64.egg/tvm/testing/aot.py”, line 843, in run_and_check return run_and_check_body(test_dir) File “/home/hex/.local/lib/python3.7/site-packages/tvm-0.14.dev156+g909c8fab8-py3.7-linux-x86_64.egg/tvm/testing/aot.py”, line 733, in run_and_check_body tvm.micro.copy_crt_config_header(“crt”, include_path) File “/home/hex/.local/lib/python3.7/site-packages/tvm-0.14.dev156+g909c8fab8-py3.7-linux-x86_64.egg/tvm/micro/build.py”, line 122, in copy_crt_config_header crt_config_path = Path(get_microtvm_template_projects(platform)) / “crt_config” / “crt_config.h” File “/home/hex/.local/lib/python3.7/site-packages/tvm-0.14.dev156+g909c8fab8-py3.7-linux-x86_64.egg/tvm/micro/build.py”, line 106, in get_microtvm_template_projects raise MicroTVMTemplateProjectNotFoundError() tvm.micro.build.MicroTVMTemplateProjectNotFoundError

Hi @alex.hex,

have you built TVM from the sources or used a precompiled version? Which TVM version are you using?


I believe it was via pip install apache-tvm may have also used the conda deploy. but compiling entirely from source with pip install for python part works nicely.

Best Regards, Alex

Haven’t tried it with pip install yet. Does this mean, you can confirm that it does not work with the pip version of tvm?