Does anyone have a copy of instruction-data-with-preference.json file?

[jingedawang]

I want to use it as input of https://github.com/rasbt/LLMs-from-scratch/blob/main/ch07/04_preference-tuning-with-dpo/dpo-from-scratch.ipynb without generating it on my machine.

Does anyone have a link of this file?

Thanks!

[rasbt]

I should be right in the GitHub folder, maybe you overlooked it 😅
https://github.com/rasbt/LLMs-from-scratch/blob/main/ch07/04_preference-tuning-with-dpo/instruction-data-with-preference.json

[jingedawang]

Oh sorry I overlooked it.

[jingedawang]

Did you share the gpt2-medium355M-sft.pth file? I didn't train it by my own. Is there anywhere to download it? Thank you!

[rasbt]

I haven't shared it because it's quite large (~1.6 GB) and can be created in ~1.98 min (depending on the GPU) with the DPO notebook.

[jingedawang]

No problem. I generated now. It's quick with Colab T4 GPU.