https://colab.research.google.com/drive/1ibPONCmdKUyjRJVcOE8fUN61mSMEbM2n?usp=sharing
Enable Long Paths via Group Policy (for Windows Pro and Enterprise users):
Press Win + R, type gpedit.msc, and press Enter to open the Group Policy Editor.
Navigate to Computer Configuration > Administrative Templates > System > Filesystem.
Double-click on the "Enable Win32 long paths" option.
Set it to "Enabled" and click "OK".
Enable Long Paths via Registry Editor (for Windows Home users):
Press Win + R, type regedit, and press Enter to open the Registry Editor.
Navigate to HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Control\FileSystem.
Find the LongPathsEnabled entry. If it doesn’t exist, create a new DWORD (32-bit) Value named LongPathsEnabled.
Set the value to 1.
Restart your computer to apply the changes.
Install NVIDIA Drivers
Ensure that you have the latest NVIDIA drivers installed. You can download and install them from the NVIDIA Driver Downloads page. Step 2: Install CUDA Toolkit
Since you want to use CUDA, you’ll need to install the CUDA Toolkit. Here’s how:
Download CUDA Toolkit:
Go to the NVIDIA CUDA Toolkit download page.
Select your operating system and version, and download the installer for CUDA 11.x (e.g., CUDA 11.8 is recommended as it is compatible with most recent PyTorch versions).
Install CUDA Toolkit:
Run the installer and follow the on-screen instructions.
During installation, make sure to install both the CUDA toolkit and the NVIDIA drivers if prompted.
Verify CUDA Installation:
After installation, you should be able to run nvcc --version in your terminal to verify the installation.
Additionally, running nvidia-smi should give you detailed information about your GPU and the installed CUDA version.
=> Run nvidia-smi in your Python terminal of your VS code
PS D:\GIT\Luxembourgish-STT> nvidia-smi
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 560.94 Driver Version: 560.94 CUDA Version: 12.6 |
|-----------------------------------------+------------------------+----------------------+
| GPU Name Driver-Model | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce GTX 960M WDDM | 00000000:01:00.0 Off | N/A |
| N/A 0C P8 N/A / 200W | 0MiB / 4096MiB | 0% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| No running processes found |
+-----------------------------------------------------------------------------------------+
Install PyTorch with CUDA Support
Once CUDA is installed, run nvcc --version then you can install PyTorch with CUDA support:
Install PyTorch with CUDA:
Install PyTorch using the following command, which will automatically select the appropriate CUDA version (replace cu118 with cu116 or cu113 if you installed a different version of CUDA):
bash
pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118
Step 4: Verify PyTorch Installation
After installation, verify that PyTorch is using CUDA:
Run a Test Script:
python
import torch
print("PyTorch version:", torch.version) print("Is CUDA available:", torch.cuda.is_available()) print("CUDA version:", torch.version.cuda)
If torch.cuda.is_available() returns True, your setup is correct, and you can use CUDA with PyTorch.