Training YoloV5 on multiple GPUs but instead it is just decreasing GPU memory usage per GPU.

[adityaee87]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

I am trying to speed up training on YoloV5.But when I am using 2 GPUs for training instead of speeding the training it is rather using half the GPU memory it was using during the training with 1 GPU. I do not know whether my fault or not because I'm submitting my job using SLURM where I ask to allocate 1 CPU and 2 gpus. Also, I have to set OMP_NUM_THREADS = 2 during training.

Additional

No response

[github-actions]

👋 Hello @adityaee87, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Requirements

Python>=3.7.0 with all requirements.txt installed including PyTorch>=1.7. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training, validation, inference, export and benchmarks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

Introducing YOLOv8 🚀

We're excited to announce the launch of our latest state-of-the-art (SOTA) object detection model for 2023 - YOLOv8 🚀!

Designed to be fast, accurate, and easy to use, YOLOv8 is an ideal choice for a wide range of object detection, image segmentation and image classification tasks. With YOLOv8, you'll be able to quickly and accurately detect objects in real-time, streamline your workflows, and achieve new levels of accuracy in your projects.

Check out our YOLOv8 Docs for details and get started with:

pip install ultralytics

[glenn-jocher]

@adityaee87 hi there! 👋

When training YOLOv5 on multiple GPUs, it is normal for each GPU to use less memory compared to training with a single GPU. This is because the model parameters are divided and processed independently on each GPU, reducing the memory required per GPU.

Regarding your SLURM job submission, it looks like you are correctly allocating both CPUs and two GPUs. However, setting OMP_NUM_THREADS=2 may not be necessary for training with YOLOv5, as this environment variable is typically used to control the number of threads in OpenMP parallel regions. YOLOv5 does not heavily utilize OpenMP, so you can try removing or adjusting this setting as needed.

If you are experiencing any other issues or have further questions, please let us know. We're here to help!

[adityaee87]

Thanks for your response @glenn-jocher but training with both 1GPU or 2GPU lead to same training speed.Is there way around it.

[glenn-jocher]

@adityaee87, thank you for your feedback. If training with both 1 GPU and 2 GPUs results in the same training speed, it could be due to other factors affecting the training process.

Here are a few suggestions to potentially improve the training speed:

1. Verify that your GPUs are properly set up and able to communicate with each other.
2. Ensure that your system has sufficient CPU power to handle multi-GPU training. Check the CPU usage during training to ensure it's not a bottleneck.
3. Review your batch size and learning rate settings. Adjusting these hyperparameters might yield better performance with multiple GPUs.
4. Check your data loading and preprocessing pipeline to ensure that it can take full advantage of multiple GPUs. Bottlenecks in data loading or augmentation can limit training speed.
5. Consider using larger models or increasing the complexity of your dataset to fully utilize the additional GPU resources.

Please let us know if you have any other questions or if there's anything else we can assist you with. We're here to help you get the most out of YOLOv5!

[adityaee87]

Thanks @glenn-jocher

[glenn-jocher]

@adityaee87 hi there! It seems that you are experiencing an issue with training YOLOv5 on multiple GPUs where the training speed remains the same regardless of using 1 or 2 GPUs.

There could be several factors contributing to this issue. Here are a few suggestions to consider:

1. Verify that your GPUs are set up correctly and can communicate with each other.
2. Ensure that your system has sufficient CPU power to handle multi-GPU training. Check the CPU usage during training to make sure it's not a bottleneck.
3. Review your batch size and learning rate settings. Adjusting these hyperparameters might improve performance with multiple GPUs.
4. Check your data loading and preprocessing pipeline to ensure it can fully utilize multiple GPUs. Bottlenecks in data loading or augmentation can limit training speed.
5. Consider using larger models or increasing dataset complexity to make better use of additional GPU resources.

Please try these suggestions and let us know if they help resolve the issue. If you have any further questions or need more assistance, feel free to ask. We're here to help you make the most of YOLOv5!