after training #961

Activity

1. jbarilo7911

   added

   enhancementNew feature or requestNew feature or request

   on Dec 28, 2024
2. UltralyticsAssistant

   added

   embeddedPertaining to IOT or low-power devicesPertaining to IOT or low-power devices

   exportsModel exports (ONNX, TensorRT, TFLite, etc.)Model exports (ONNX, TensorRT, TFLite, etc.)

   on Dec 28, 2024
3. UltralyticsAssistant commented on Dec 28, 2024

   

   UltralyticsAssistant

   on Dec 28, 2024

   Member

   More actions

   👋 Hello @jbarilo7911, thank you for raising an issue about Ultralytics HUB 🚀! It looks like you're suggesting the addition of export-related optimizations and model pruning capabilities, particularly for edge devices and formats like TFLite. This is a great suggestion and something we understand could be valuable for many users.

   If this is a feature request, we appreciate the idea and will take it into consideration for future developments. If there are any specific scenarios or requirements you'd like this feature to address, feel free to share more details!

   If this is a 🐛 Bug Report, please also include detailed steps to reproduce the issue to help us investigate.

   We aim to address all issues in a timely manner, and an Ultralytics engineer will review this soon. Thanks for your patience! 😊
4. pderrenger commented on Dec 28, 2024

   

   pderrenger

   on Dec 28, 2024

   Member

   More actions

   @jbarilo7911 thank you for your suggestion and for taking the time to share your thoughts! 😊

   The features you're requesting—adding an optimize=vela option during export and incorporating model pruning capabilities—are excellent ideas, especially for improving deployment on resource-constrained edge devices. Here's some context and guidance:

   1. Optimize with Vela:
      Vela is a compiler specifically designed to optimize TensorFlow Lite models for Arm Ethos-U NPUs. While Vela isn't natively integrated into the current Ultralytics export workflow, you can follow a two-step process:

      • First, export your YOLO model to TFLite using the model.export(format='tflite') function.
      • Then, optimize the exported TFLite model using the Vela compiler. You can find more details on Vela in Arm's documentation.

      We'll forward your suggestion to the development team for consideration as part of future updates to simplify this process directly in the HUB.

   2. Model Pruning:
      While pruning isn't directly integrated into HUB yet, it's a well-known technique to reduce model size and improve inference speed, particularly for edge devices. For now, you can manually prune models using third-party libraries like TensorFlow Model Optimization Toolkit or PyTorch's pruning utilities before exporting to TFLite or other formats. For example:

      from torch.nn.utils import prune
      model = YOLO('path/to/best.pt')
      # Apply pruning
      model.model = prune.global_unstructured(
          parameters=model.model.parameters_to_prune,
          pruning_method=prune.L1Unstructured,
          amount=0.5,  # Prune 50% of parameters
      )
      # Export pruned model
      model.export(format='tflite')

      We'll also pass your request for built-in pruning support to the team, as this feature could be incredibly beneficial for edge deployments.

   If you're interested in learning more about the current export workflows or supported optimizations, feel free to check out the Export Documentation and Edge TPU guide.

   Thanks again for the valuable feedback—your input helps us refine and enhance HUB for everyone! Let us know if you have any additional questions or ideas. 🚀
5. github-actions commented on Nov 23, 2025

   

   github-actionsbot

   on Nov 23, 2025 – with GitHub Actions

   More actions

   👋 Hello there! We wanted to give you a friendly reminder that this issue has not had any recent activity and may be closed soon, but don't worry - you can always reopen it if needed. If you still have any questions or concerns, please feel free to let us know how we can help.

   For additional resources and information, please see the links below:

   • Docs: https://docs.ultralytics.com
   • HUB: https://hub.ultralytics.com
   • Community: https://community.ultralytics.com

   Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

   Thank you for your contributions to YOLO 🚀 and Vision AI ⭐
6. github-actions

   added

   StaleStale and schedule for closing soonStale and schedule for closing soon

   on Nov 23, 2025