[Bug] Debug version run failed
- Dominant language
- Python
- Stars
- 8.1k
- Forks
- 748
- Avg merge
- 6d 2h
- Merged PRs (30d)
- 54
Description
### Checklist
- [x] 1. I have searched related issues but cannot get the expected help.
- [X] 2. The bug has not been fixed in the latest version.
### Describe the bug
```
[TM][DEBUG] void turbomind::cublasMMWrapper::Gemm(cublasOperation_t, cublasOperation_t, int, int, int, const void*, int, const void*, int, void*, int)
[TM][DEBUG] void turbomind::cublasMMWrapper::Gemm(cublasOperation_t, cublasOperation_t, int, int, int, const void*, int, const void*, int, void*, int, float, float)
terminate called after throwing an instance of 'std::runtime_error'
what(): [TM][ERROR] CUDA runtime error: misaligned address /data/lmdeploy/lmdeploy/src/turbomind/utils/cublasMMWrapper.cc:303
```
I compiled the debug version, and replace the so file.
```
cp ./lib/_turbomind.cpython-38-x86_64-linux-gnu.so /opt/conda/lib/python3.8/site-packages/lmdeploy/lib/_turbomind.cpython
-38-x86_64-linux-gnu.so
```
But the run failed.
Any advice? thank you!
### Reproduction
debug version
```
#!/bin/sh
cmake .. \
-DCMAKE_BUILD_TYPE=Debug \
-DCMAKE_EXPORT_COMPILE_COMMANDS=1 \
-DCMAKE_INSTALL_PREFIX=./install \
-DBUILD_PY_FFI=ON \
-DBUILD_MULTI_GPU=ON \
-DCMAKE_CUDA_FLAGS="-lineinfo" \
-DUSE_NVTX=ON \
-DSM=70
```
### Error traceback
_No response_
Contributor guide
Assessment
This issue has not been assessed yet.