Interest in Your Research and a Request for Data Access

[Wangjiangxue]

I am an undergraduate student currently studying your papers, and I find your work truly inspiring. I am particularly interested in your code and eager to understand how it was implemented. To fully explore it, I would need access to the specific dataset you used. If possible, could you kindly share the data or direct me to the website where I can obtain it? I would be sincerely grateful for your guidance.

[Xiaoyu-Tao]

Hi, thanks for your interest! 😊
You can download the dataset from the following Google Drive link:
🔗 https://drive.google.com/drive/folders/19P---oV4nQ53JgKnE0VX3t_N1jLliVSv

Please let me know if you encounter any access issues or missing files.

[Wangjiangxue]

Thank you so much!

[Sample-design-alt]

Thanks for your share! But i can't found the rwc fd ecg datasets in the link

[zhjai]

Thanks for the follow-up! The remaining datasets are available on Hugging Face:
https://huggingface.co/datasets/zhjai/InstructTime/tree/main
Feel free to download and let us know here if you need anything else.

[Sample-design-alt]

Thanks for your help! Actually, I want to reproduce the Instructime in the 30 UEA datasets. do you have some results on UEA?
OR How should I do? Can you give some sugguestion to help reproduce?
Based on the code you provided, I think I need to build a tokenizer for each dataset before proceeding with joint downstream fine-tuning.

[zhjai]

Thanks for your interest in InstructTime! We don’t have a full result table on all 30 UEA datasets to share yet, but your understanding is right: you need to build a tokenizer (using our tokenizer scripts) for each dataset first, and then run the joint downstream fine-tuning. For UEA datasets, I recommend keep a small learning rate for stable training.
Hope this helps, and feel free to open another issue if anything is unclear.