Hello, I am trying to use my GPU for training, so I downloaded CUDA and cudnn, I copied all the files from the cudnn file to CUDA, and I added the bin, libnvvp, and extras\cupti\lib64 to paths in the environment variable
the virtual environment I am trying to run model_builder_tf2_test.py, and it gives me a warning like this dlerror: cudart64_110.dll not found , I checked the file it is in the NVIDIA GPU Computing Toolkit\CUDA\v11.0\bin folder but it is not seeing for some reason
I need to work with GPU so please if the answer to this I am open to it
Welcome to the Tensorflow Forum!
Could you please elaborate on how you installed the Tensorflow package? Please also include the output of the following code:
python3 -c "import tensorflow as tf; print(tf.config.list_physical_devices('GPU'))"
nvidia-smi
Thank you!
I download the TensorFlow throw Powershell in visual studio 2019, The output when I apply the command line in the PowerShell is
âC:\Users\osama\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.9_qbz5n2kfra8p0\LocalCache\local-packages\Python39\site-packages\tensorflow_init_.pyâ, line 37, in
from tensorflow.python.tools import module_util as _module_util
ModuleNotFoundError: No module named âtensorflow.pythonâ
It seems Tensorflow library is not installed properly on your machine.
Following are the the possible reasons in your case:
1. You need to install the MSVC 2019 redistributable
2. Your CPU does not support AVX2 instructions
3. Your CPU/Python is on 32 bits
Please check and let us know ?
Thank you!
i tried to use another IDE , but now I am using VS code , I checked to also the AVX my system is supporting AVX, and then I changed the Python to 64-bit the command up there is start to give me like this
(tfod) PS C:\Users*****\source\repos\TFODCourse> python3 -c âimport tensorflow as tf; print(tf.config.list_physical_devices(âGPUâ))â
2023-06-08 13:59:50.881271: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcudart64_110.dllâ; dlerror: cudart64_110.dll not found
2023-06-08 13:59:50.881698: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.
2023-06-08 13:59:55.827809: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcudart64_110.dllâ; dlerror: cudart64_110.dll not found
2023-06-08 13:59:55.828280: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcublas64_11.dllâ; dlerror: cublas64_11.dll not found
2023-06-08 13:59:55.828822: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcublasLt64_11.dllâ; dlerror: cublasLt64_11.dll not found
2023-06-08 13:59:55.829396: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcufft64_10.dllâ; dlerror: cufft64_10.dll not found
2023-06-08 13:59:55.829918: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcurand64_10.dllâ; dlerror: curand64_10.dll not found
2023-06-08 13:59:55.830454: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcusolver64_11.dllâ; dlerror: cusolver64_11.dll not found
2023-06-08 13:59:55.830956: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcusparse64_11.dllâ; dlerror: cusparse64_11.dll not found
2023-06-08 13:59:55.831428: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library âcudnn64_8.dllâ; dlerror: cudnn64_8.dll not found
2023-06-08 13:59:55.831567: W tensorflow/core/common_runtime/gpu/gpu_device.cc:1934] Cannot dlopen some GPU libraries. Please make sure the missing libraries mentioned above are installed properly if you would like to use GPU. Follow the guide at Install TensorFlow with pip for how to download and setup the required libraries for your platform.
Skipping registering GPU devicesâŚ
The hoops the users have to go through just to install tensorflow gpu are monumental.
I hope you guys get to fix the issues.
The only reason I need TF is so that I can reproduce github code for my work. Otherwise, Torch is so much easier.
TF slows down everyoneâs work
1 Like