How to train yolov8 using intel arc 770 gpu?

[johnkim7]

I bought an Intel Arc 770 with a 13th gen CPU desktop to use for training the YOLOv8 model. However, I couldn't find a way to use it. There is an option for CUDA, but not for the Arc 770.

@https://docs.ultralytics.com/modes/train/#usage-examples

I am using a Python Jupyter notebook. Can anyone help with this?

[ACarmona_Intel]

Hello Johnkim7,

Thank you for posting in our communities. 

We will do further research on this matter and post the response on this thread once it is available.

Thank you for your patience and understanding!

Best regards,

CarmonaA.

Intel Customer Support Technician

[Siyabonga · Accepted solution]

Using an Intel Arc GPU, such as the Arc 770, for training machine learning models like YOLOv8 in a Python Jupyter notebook can be challenging, particularly because most popular deep learning frameworks, such as TensorFlow and PyTorch, are optimized for NVIDIA GPUs using CUDA. Intel has its own set of tools and libraries for GPU computing, and integrating these into your workflow will require using Intel's software stack. Here’s how you can proceed:

1. Install Intel OneAPI Toolkits

Intel OneAPI provides a comprehensive set of tools for data-centric workloads. For using Intel GPUs, you'll need the Intel OneAPI Base Toolkit and the Intel OneAPI AI Analytics Toolkit.

1. Download and Install Intel OneAPI Toolkits:

   • Go to the Intel OneAPI Toolkits page and download the Base Toolkit and AI Analytics Toolkit.
   • Follow the installation instructions provided on the site.

2. Set Up the Environment:

   • After installation, set up the environment by sourcing the setvars.sh script. This script is usually located in the installation directory.
     bash

     Copy code

     source /opt/intel/oneapi/setvars.sh

2. Install Intel Extension for PyTorch

Intel provides an extension for PyTorch to enable optimized performance on Intel hardware.

1. Install the Extension:

   • You can install the Intel Extension for PyTorch using pip:
     bash

     Copy code

     pip install intel-extension-for-pytorch

2. Modify Your PyTorch Code:

   • To use Intel hardware, you need to import the extension and set the appropriate device.
     python

     Copy code

     import torch import intel_extension_for_pytorch as ipex device = torch.device('xpu' if torch.xpu.is_available() else 'cpu') # Your model definition model = YourModel().to(device) # Optimizer and loss function optimizer = torch.optim.Adam(model.parameters()) loss_fn = torch.nn.CrossEntropyLoss() # Training loop for epoch in range(num_epochs): for data, target in train_loader: data, target = data.to(device), target.to(device) optimizer.zero_grad() output = model(data) loss = loss_fn(output, target) loss.backward() optimizer.step()

3. Training YOLOv8

If you are specifically using YOLOv8, which is implemented using PyTorch, you can adapt the example code above to ensure that it utilizes Intel's hardware.

1. Install YOLOv8 and Dependencies:

   • Ensure you have the required dependencies installed:
     bash

     Copy code

     pip install ultralytics

2. Modify YOLOv8 Code to Use Intel GPU:

   • Adapt the YOLOv8 training script to utilize the Intel GPU.
     python

     Copy code

     from ultralytics import YOLO import torch import intel_extension_for_pytorch as ipex # Check for Intel GPU availability device = torch.device('xpu' if torch.xpu.is_available() else 'cpu') # Load the YOLOv8 model model = YOLO('yolov8.yaml').to(device) # Train the model model.train(data='path/to/dataset', epochs=50, device=device)

4. Run Jupyter Notebook with Intel GPU Support

When running your Jupyter Notebook, ensure that the environment variables and paths are correctly set up to utilize the Intel GPU.

1. Start Jupyter Notebook:

   • Open a terminal, activate the OneAPI environment, and start Jupyter Notebook:
     bash

     source /opt/intel/oneapi/setvars.sh jupyter notebook

2. Run Your Notebook:

   • Ensure your notebook contains the modified code to leverage Intel's GPU capabilities as described above.

Troubleshooting

• Check Compatibility: Make sure your Intel Arc GPU is supported by the version of Intel OneAPI and Intel Extension for PyTorch you are using.
• Update Drivers: Ensure your system has the latest Intel GPU drivers installed.
• Consult Documentation: Refer to the Intel OneAPI documentation for more detailed guidance and troubleshooting steps.

By following these steps, you should be able to leverage your Intel Arc 770 GPU for training YOLOv8 models in a Python Jupyter notebook.

[johnkim7]

Hello Siyabonga,

Thanks to your guidance, I achieved a great result.

Thanks a lot.

Here is the screen capture of what I did. Hurray!

[ACarmona_Intel]

Hello Johnkim7,

We wanted to inform you that it's up to the developers to choose whether their application or software supports Intel ARC or not. In regards to this, we recommend that you contact the application or software developer directly for support.

However, a member(@Siyabonga) of our community has addressed your issue; kindly review it and follow the steps suggested to see if they fix the problem.

By the way, for additional information, please submit a new question, as this thread will no longer be monitored.

Best regards,

Carmona A.

Intel Customer Support Technician