-
Notifications
You must be signed in to change notification settings - Fork 1.5k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Update unsloth_integration.md #2742
base: main
Are you sure you want to change the base?
Conversation
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks!! I left a few comments to iterate
docs/source/unsloth_integration.md
Outdated
|
||
Section under construction. Feel free to contribute! | ||
## Table of Contents | ||
- [Installation](#installation) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
No need for a toc, you will have one automatically on the right with the doc builder
docs/source/unsloth_integration.md
Outdated
3. **Install Optional Dependencies**: | ||
```bash | ||
pip install flash-attn==2.5.8 trl==0.8.6 accelerate==0.27.2 |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Why do you need this very old version of trl?
from transformers import TrainingArguments | ||
|
||
trainer = TrainingArguments( |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Trainer and training args are different things. There's probably an issue here
docs/source/unsloth_integration.md
Outdated
- Python 3.8+ | ||
- PyTorch 2.0+ | ||
- NVIDIA GPU with FP16 support (recommended) |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should probably also specify it is limited to single GPU
What does this PR do?
Fixes # (issue)
Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.