ultralytics / hub

Ultralytics HUB tutorials and support
https://hub.ultralytics.com
GNU Affero General Public License v3.0
138 stars 14 forks source link

TypeError : unhashable type: 'numpy.ndarray' #845

Closed TepongLekTholang closed 1 month ago

TepongLekTholang commented 2 months ago

Search before asking

HUB Component

Training, Export

Bug

WhatsApp Image 2024-09-19 at 10 58 19_9f2b339d I have annotated 1,000 images in Roboflow (using instance segmentation), and when I tried to train the data using Google Colab with the preset snippet from Roboflow, it didn't work due to an error about an array. I looked for solutions through Gemini, GPT, Blackbox, Google, Stack Overflow, and GitHub, changing/adding/installing various things, but in the end, I still couldn't resolve it. I’ve been stuck here for four days, and I really need this to meet my project deadline next week. I also asked my friends who usually use Roboflow, and they suggested using YOLOv10. However, for some reason, they have the option to "Train with YOLOv10" on their accounts, while I only have YOLOv5 and YOLOv8. Please, your help would mean a lot to me and my career. Thank you!

WhatsApp Image 2024-09-19 at 14 54 03_6290fb19

Environment

No response

Minimal Reproducible Example

No response

Additional

No response

UltralyticsAssistant commented 2 months ago

👋 Hello @TepongLekTholang, thank you for raising an issue about the Ultralytics HUB repository 🚀!

We're sorry to hear about the trouble you're experiencing with training your dataset. Here are some helpful resources that might assist you:

For the TypeError you're encountering, providing a minimum reproducible example (MRE) can greatly assist us in identifying the issue. You can find guidance on creating an MRE here.

We advise including any screenshots and the steps you followed, which can further help in getting to the bottom of this bug 🐛.

An Ultralytics engineer will also review your issue soon and provide further assistance. We appreciate your patience and understanding. Thank you!

TepongLekTholang commented 2 months ago

Hello, I just followed the advice you gave! I exported the Roboflow dataset and selected the Ultralytics HUB format so I could train the data through Ultralytics HUB, but why is there no option for that?

Screenshot 2024-09-23 090610

glenn-jocher commented 2 months ago

@TepongLekTholang hello! 😊

Thanks for following up. It sounds like you're on the right track by exporting your dataset in the Ultralytics HUB format. If you're not seeing the option to train directly in the HUB, here are a few steps to help you proceed:

  1. Verify Dataset Upload: Ensure your dataset is correctly uploaded to the Ultralytics HUB. You can do this by checking the Datasets section in your HUB account.

  2. Create a Project: If you haven't already, create a new project in the HUB. This will allow you to organize and manage your dataset and models effectively.

  3. Train a Model: Once your dataset is uploaded and your project is set up, you should be able to select your dataset and start training a model from the Models section.

If you're still encountering issues, please ensure you're using the latest version of the Ultralytics HUB. Sometimes updates can resolve unexpected behavior.

Feel free to reach out if you have any more questions or need further assistance. We're here to help! 🚀

TepongLekTholang commented 2 months ago

I did it! thanks guys so i went to Roboflow> version menu> download zip as Yolov8> upload dataset (recent downloaded zip) to ultralytics hub by creating new project> models menu> train> choose google colab> copy snippet> that's it

glenn-jocher commented 1 month ago

Hello! 😊

That's fantastic to hear! 🎉 I'm glad you were able to successfully upload and train your dataset using the Ultralytics HUB. Your detailed steps will surely help others facing similar challenges.

If you have any more questions or need further assistance, feel free to reach out. We're here to support you in your YOLO journey. Happy modeling! 🚀

And remember, the YOLO community and the Ultralytics team are always here to help.