The warning “Could not find the Cudart64_90.dll” has been reported by numerous users when using TensorFlow or comparable programs. For developers, this problem can be especially annoying because it can make machine learning projects and other CUDA-reliant applications less efficient.
Fortunately, there are workable ways to deal with this issue. We’ll walk you through the process today to fully fix it so you can resume creating without any more disruptions.
Table of Contents
What Should I Do if There is No Cudart64_90.dll?
Here are some ways which can help you to solve the issue of Cudart64_90.dll is missing:
Method 1: Install CUDA Toolkit 9
- Go to the download page of CUDA Toolkit 9
- Choose your operating system and the version
- To download the Base Installer, click on Download in the Base Installer section.
- Run the setup file after downloading, then adhere to the directions.
Try both versions; some users claim that the offline version is more effective than the network installer, while others claim that the offline version is superior.
Method 2: Copy the Missing Files Manually
- Extract CUDA Toolkit 9 after downloading it. The setup file should not be run.
- After using the Windows key + R, type %localappdata%.
- Next, go to
Temp\CUDA\cudart\bin
- Look for and copy cudart64_90.dll there.
- Next, go to
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v9.1\bin
- The DLL should be put in that directory.
Be aware that the file path in Step 5 may vary based on the CUDA version you have installed.
Method 3: Reinstall CUDA Toolkit 9
- In order to access Apps and Features, use the Windows key + X.
- Find CUDA Toolkit 9, choose Uninstall by clicking the three dots next to it.
- Observe the guidelines displayed on the screen.
- Download CUDA Toolkit 9 once more after removing it.
For this solution to function, you must uninstall any other installed versions, such as 9.1.
Method 4: Change PATH variables
- Verify that CUDA and cuDNN are installed. For this solution to function, version 9 is required.
- Type “view advanced” after pressing the Windows key + S. From the list, select View advanced system settings.
- On Environment Variables, click.
- Double-click your user’s Path variable, then add CUDA and CuDNN paths that contain cudart64_90.dll.
- Save your modifications.
- Add the same paths from Step 4 to the Path variable in the System variables.
- Save your modifications.
We appreciate you reading! For additional resources and troubleshooting advice, visit Trendblog if our tutorial on fixing the Cudart64_90.dll issue was useful. Maintain your knowledge and ensure the success of your machine learning initiatives!