FLAN-T5-small model run sentiments.ipynb demo gives so so many ""using bos_token ,but it is not set yet" lines"
Open
bug
- Dominant language
- Python
- Stars
- 4.8k
- Forks
- 487
- PR merge metrics
- No merged PRs in 30d
Description
### 🐛 Describe the bug
when I try to replace GPT2 with FLAN-T5-small model to run this demo:
https://colab.research.google.com/github/CarperAI/trlx/blob/main/examples/notebooks/trlx_sentiments.ipynb
it output a lot of "using bos_token ,but it is not set yet" lines
and I find when I set the samples smaller, like smaples = imdb['text'][:20]
the lines number is just 20
so ..when I use the whole imdb dataset ,the output will be 50k lines.
And before its final 50k-output done ,nothing else will be seen.
### Which trlX version are you using?
latest at 20230416
### Additional system and package information
colab Python env py39
Contributor guide
Assessment
This issue has not been assessed yet.