Description
Description
This library appears to use an old version of the CUDA toolkit libraries. The readme on this project doesn't document which release I need to install so that I can use this library with the GPU. It isn't apparent from the file names, or list of releases on NVidia's archive what version I need to install.
For example, one error indicates Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found. When I go into the install path under the environment variable CUDA_PATH that was set during installation there is a file bin\cudart64_12.dll (along with some others). It looks like it's checking for different versions of the same files.
I have the latest versions of CUDA toolkit, cuDNN, and NVidia drivers installed. All the libraries needed to run tensorflow should be there. If there's a specific version I need to download, please let me know. It should be documented in the Installation section of the readme.
Reproduction Steps
- Made a new project and read out sample data.
- Installed the two most recent versions of CUDA toolkit (10 and 11) from: https://developer.nvidia.com/cuda-downloads?target_os=Windows&target_arch=x86_64
Example:
using keras = Tensorflow.Keras;
using Tensorflow.Keras.Datasets;
public class Brain
{
public DatasetPass Dataset = new keras.Datasets.Mnist().load_data();
}
Then I displayed the data in another file: infoLabel.Text = $"{nameof(Tensorflow.NumPy.NDArray)} = {brain.Dataset.Train.Item1}";
Received the following error messages in the debugger:
2025-02-16 17:50:34.898721: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2025-02-16 17:50:34.899170: 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.
...
2025-02-16 17:50:35.145616: I tensorflow/core/platform/cpu_feature_guard.cc:193] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations: AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
2025-02-16 17:50:35.208050: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cudart64_110.dll'; dlerror: cudart64_110.dll not found
2025-02-16 17:50:35.209043: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cublas64_11.dll'; dlerror: cublas64_11.dll not found
2025-02-16 17:50:35.209730: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cublasLt64_11.dll'; dlerror: cublasLt64_11.dll not found
2025-02-16 17:50:35.210525: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cufft64_10.dll'; dlerror: cufft64_10.dll not found
2025-02-16 17:50:45.067464: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cusparse64_11.dll'; dlerror: cusparse64_11.dll not found
2025-02-16 17:50:45.068052: W tensorflow/stream_executor/platform/default/dso_loader.cc:64] Could not load dynamic library 'cudnn64_8.dll'; dlerror: cudnn64_8.dll not found
2025-02-16 17:50:45.068074: 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 https://www.tensorflow.org/install/gpu for how to download and setup the required libraries for your platform.
Skipping registering GPU devices...
Known Workarounds
No response
Configuration and Other Information
<TargetFramework>net9.0</TargetFramework>
...
<ItemGroup>
<PackageReference Include="SciSharp.TensorFlow.Redist-Windows-GPU" Version="2.10.3" />
<PackageReference Include="TensorFlow.Keras" Version="0.15.0" />
<PackageReference Include="TensorFlow.NET" Version="0.150.0" />
</ItemGroup>