minor fixes

[d-kleine]

• fixed typos
• fixed formatting
• changed path call for better convenience

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB

[d-kleine]

For reproducibility, you might add a context length to the model params (as in the instruction finetuning nb)

[rasbt]

The fixes look good to me. Much appreciated!

[rasbt]

For reproducibility, you might add a context length to the model params (as in the instruction finetuning nb)

I originally had that, and was thinking about that, but I decided to remove it because it could lead to artificial truncation if someone uses it on a larger datasets with longer responses.

[d-kleine]

Ah, I see - thanks!

[d-kleine]

fyi ollama/ollama#6060

[d-kleine]

@rasbt Llama 3.1 models have been updated some days ago (due to issue with the tokenizer iirc), updates have now been pushed to the Ollama models too:
https://ollama.com/library/llama3.1

Might be relevant for this nb (ch07/04_preference-tuning-with-dpo/create-preference-data-ollama.ipynb)

You can update it with ollama run llama3.1:70b, you might need to update Ollama before.

[rasbt]

Thanks for the note! I think given that it doesn't change any of the settings, I'd leave it as is for now, but it's good to know!