QuentinFuxa
Cannot load symbol cudnnCreateTensorDescriptor
Im using CachyOS linux with GNOME 49, not sure if i missed a step or did something wrong, using a python virtual environment
transcribe ❯ whisperlivekit-server --host 0.0.0.0 --port 80 --model small.en INFO: Started server process [114576] INFO: Waiting for application startup. Simulstreaming will use Faster Whisper for the encoder. Unable to load any of {libcudnn_ops.so.9.1.0, libcudnn_ops.so.9.1, libcudnn_ops.so.9, libcudnn_ops.so} Invalid handle. Cannot load symbol cudnnCreateTensorDescriptor fish: Job 1, 'whisperlivekit-server --host 0.…' terminated by signal SIGABRT (Abort)
Hi, this error means your system can’t find a compatible cuDNN library. Faster-Whisper requires that your CUDA + cuDNN versions match the PyTorch/CTranslate2 builds you've installed.
On Arch/CachyOS the easiest fix is usually:
sudo pacman -S cuda cudnn
sudo ldconfig
Then ensure the cuDNN .so files are visible:
ls /usr/lib/libcudnn*
If you installed PyTorch via pip, verify the CUDA version it expects:
python - <<EOF
import torch
print(torch.version.cuda)
EOF
Match that with your installed CUDA:
nvcc --version
Hi, this error means your system can’t find a compatible cuDNN library. Faster-Whisper requires that your CUDA + cuDNN versions match the PyTorch/CTranslate2 builds you've installed.
On Arch/CachyOS the easiest fix is usually:
sudo pacman -S cuda cudnn sudo ldconfig Then ensure the cuDNN
.sofiles are visible:ls /usr/lib/libcudnn* If you installed PyTorch via
pip, verify the CUDA version it expects:python - <<EOF import torch print(torch.version.cuda) EOF Match that with your installed CUDA:
nvcc --version
thank you it works now, i do have a strange error with it though, sometimes the text it shows will bounce up and down on the screen really fast. Also i tried --diarization but got errors with that
`transcribe ❯ pip install nemo_toolkit['asr']
Building wheel for ml_dtypes (pyproject.toml) ... error error: subprocess-exited-with-error
× Building wheel for ml_dtypes (pyproject.toml) did not run successfully. │ exit code: 1 ╰─> [23 lines of output] running bdist_wheel running build running build_py creating build creating build/lib.linux-x86_64-cpython-313 creating build/lib.linux-x86_64-cpython-313/ml_dtypes copying ml_dtypes/_iinfo.py -> build/lib.linux-x86_64-cpython-313/ml_dtypes copying ml_dtypes/_finfo.py -> build/lib.linux-x86_64-cpython-313/ml_dtypes copying ml_dtypes/init.py -> build/lib.linux-x86_64-cpython-313/ml_dtypes copying ml_dtypes/py.typed -> build/lib.linux-x86_64-cpython-313/ml_dtypes running build_ext building 'ml_dtypes._ml_dtypes_ext' extension creating build/temp.linux-x86_64-cpython-313 creating build/temp.linux-x86_64-cpython-313/ml_dtypes creating build/temp.linux-x86_64-cpython-313/ml_dtypes/_src gcc -fno-strict-overflow -Wsign-compare -DNDEBUG -g -O3 -Wall -march=native -O3 -pipe -fno-plt -fexceptions -Wp,-D_FORTIFY_SOURCE=3 -Wformat -Werror=format-security -fstack-clash-protection -fcf-protection -mpclmul -fno-semantic-interposition -g -ffile-prefix-map=/tmp/pkg/src=/usr/src/debug/python -flto=auto -ffat-lto-objects -fno-semantic-interposition -march=native -O3 -pipe -fno-plt -fexceptions -Wp,-D_FORTIFY_SOURCE=3 -Wformat -Werror=format-security -fstack-clash-protection -fcf-protection -mpclmul -fno-semantic-interposition -g -ffile-prefix-map=/tmp/pkg/src=/usr/src/debug/python -flto=auto -march=native -O3 -pipe -fno-plt -fexceptions -Wp,-D_FORTIFY_SOURCE=3 -Wformat -Werror=format-security -fstack-clash-protection -fcf-protection -mpclmul -fno-semantic-interposition -g -ffile-prefix-map=/tmp/pkg/src=/usr/src/debug/python -flto=auto -fPIC -Ithird_party/eigen -Iml_dtypes -I/tmp/pip-build-env-t9ljh83i/overlay/lib/python3.13/site-packages/numpy/_core/include -I/home/josepho/transcribe/include -I/usr/include/python3.13 -c ml_dtypes/_src/dtypes.cc -o build/temp.linux-x86_64-cpython-313/ml_dtypes/_src/dtypes.o -std=c++17 -DEIGEN_MPL2_ONLY -fvisibility=hidden -ftrapping-math In file included from third_party/eigen/Eigen/Core:205, from ml_dtypes/_src/dtypes.cc:32: third_party/eigen/Eigen/src/Core/arch/AVX512/TrsmKernel.h:51:10: fatal error: TrsmUnrolls.inc: No such file or directory 51 | #include "TrsmUnrolls.inc" | ^~~~~~~~~~~~~~~~~ compilation terminated. error: command '/usr/bin/gcc' failed with exit code 1 [end of output]
note: This error originates from a subprocess, and is likely not a problem with pip. ERROR: Failed building wheel for ml_dtypes Building wheel for editdistance (pyproject.toml) ... done Created wheel for editdistance: filename=editdistance-0.8.1-cp313-cp313-manylinux_2_42_x86_64.whl size=395974 sha256=de0635a6262839e0a6f2f4e29208c0fce554cfcf321aaec92040704d1d684d4e Stored in directory: /home/josepho/.cache/pip/wheels/5d/f0/83/35579f35f97d5b90384045f587f969100fd0ee02b80fe95317 Building wheel for kaldi-python-io (pyproject.toml) ... done Created wheel for kaldi-python-io: filename=kaldi_python_io-1.2.2-py3-none-any.whl size=9008 sha256=710096ada26721f2fafeb7b49083fb3547c05592745e7f023d1772c632e1bf72 Stored in directory: /home/josepho/.cache/pip/wheels/5c/93/62/04ab2b6586d42b43a54b97107fca2a3448f645e2e6c4c5e6eb Building wheel for docopt (pyproject.toml) ... done Created wheel for docopt: filename=docopt-0.6.2-py2.py3-none-any.whl size=13812 sha256=616d6d73508ce58a2ffd4c7302f1ebf882c58ed177e24da86315bb3187c6706c Stored in directory: /home/josepho/.cache/pip/wheels/0b/1d/03/175286677fb5a1341cc3e4755bf8ec0ed08f3329afd67446b0 Building wheel for wget (pyproject.toml) ... done Created wheel for wget: filename=wget-3.2-py3-none-any.whl size=9667 sha256=a44515064b8a63375665f8ea415a63bc6f3f7a60868aed3ec8acb76abe950911 Stored in directory: /home/josepho/.cache/pip/wheels/8a/b8/04/0c88fb22489b0c049bee4e977c5689c7fe597d6c4b0e7d0b6a Successfully built nemo_toolkit ctc_segmentation antlr4-python3-runtime numpy sox intervaltree editdistance kaldi-python-io docopt wget Failed to build ml_dtypes error: failed-wheel-build-for-install
× Failed to build installable wheels for some pyproject.toml based projects ╰─> ml_dtypes`