YouTipPytorch Torch Cuda Is_Available
## PyTorch: torch.cuda.is_available() Reference
In PyTorch, `torch.cuda.is_available()` is a fundamental utility function used to determine whether your current system environment supports CUDA (Compute Unified Device Architecture).
CUDA is a parallel computing platform and application programming interface (API) model created by NVIDIA. By checking if CUDA is available, you can dynamically write code that leverages GPU acceleration when a compatible NVIDIA graphics card and the correct drivers are installed, while gracefully falling back to the CPU when they are not.
---
### Function Definition
```python
torch.cuda.is_available()
```
#### Return Value
* **`bool`**: Returns `True` if a CUDA-supporting GPU is detected and the PyTorch CUDA library is correctly configured; otherwise, returns `False`.
---
### Basic Usage and Code Example
The most common design pattern in PyTorch is to check for CUDA availability at the beginning of your script to define a target `device`. This device object is then used to cast tensors and neural network models to either the GPU or CPU.
```python
import torch
# 1. Check if CUDA is available
cuda_present = torch.cuda.is_available()
print(f"Is CUDA available? {cuda_present}")
# 2. Define the target device dynamically
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
print(f"Using device: {device}")
# 3. Retrieve additional GPU information if available
if torch.cuda.is_available():
print("CUDA is available!")
print("Number of GPUs:", torch.cuda.device_count())
print("Primary GPU Name:", torch.cuda.get_device_name(0))
else:
print("CUDA is not available. Running on CPU.")
```
---
### Practical Workflow: Moving Tensors and Models
Once you have verified CUDA availability and defined your `device`, you can seamlessly move your PyTorch operations to the GPU.
```python
import torch
# Define device
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# Create a tensor on the host (CPU)
x = torch.tensor([1.0, 2.0, 3.0])
# Move the tensor to the selected device (GPU if available, otherwise CPU)
x = x.to(device)
print(f"Tensor device: {x.device}")
# Define a simple model and move it to the device
model = torch.nn.Linear(3, 1).to(device)
print(f"Model parameters device: {next(model.parameters()).device}")
```
---
### Troubleshooting: Why does `torch.cuda.is_available()` return `False`?
If you expect your GPU to be active but the function returns `False`, check the following common issues:
1. **Missing or Incompatible NVIDIA Drivers**: Ensure that your system has the official NVIDIA proprietary drivers installed and updated to a version that supports your CUDA toolkit.
2. **CPU-only PyTorch Installation**: If you installed PyTorch via a simple `pip install torch` command without specifying the CUDA variant, you might have downloaded the CPU-only version. Verify your installation command on the (https://pytorch.org/).
3. **CUDA Toolkit Version Mismatch**: The CUDA version supported by your installed PyTorch build must be compatible with the CUDA version installed on your system. You can check your PyTorch-compiled CUDA version using:
```python
print(torch.version.cuda)
```
4. **No NVIDIA GPU Present**: CUDA is exclusive to NVIDIA hardware. If you are running on an AMD GPU, Intel GPU, or Apple Silicon (M1/M2/M3), `torch.cuda.is_available()` will return `False`. *(Note: For Apple Silicon, use `torch.backends.mps.is_available()` instead).*