Hi, I am new to TVM. I just built TVM from source in a clean nvidia-docker container with tag:11.8.0-cudnn8-devel-ubuntu22.04. I set set(USE_CUDA ON) and set(USE_LLVM OFF) because I am using cuda backend. Then I tried to follow the tutorial here.
Now in python, tvm seems to work well:
>>> import tvm
>>> tvm.cuda().exist
> True
Then there appear two problems when I directly run python tune_relay_cuda.py.
First, it tells me Check failed: (bf != nullptr) is false: target.build.llvm is not enabled. Why is LLVM still required in this codegen with cuda target?
Second, it tells me FileNotFoundError: [Errno 2] No such file or directory: 'resnet-18.log.tmp'. It is reasonable because in the tune_tasks function the tmp log file is never created. There must be something I have missed.
Can anyone give me any advice?