Skip to content

[docs] Clarify dtype defaults between trf v5 and TRL #5457

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.

Sign up for GitHub

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

Open
casinca wants to merge 1 commit into
base: main
Choose a base branch
from
+7 −0
Open

[docs] Clarify dtype defaults between trf v5 and TRL #5457

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
7 changes: 7 additions & 0 deletions docs/source/customization.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,13 @@ TRL is designed with modularity in mind so that users are able to efficiently cu
> [!NOTE]
> Although these examples use the [`DPOTrainer`], these customization methods apply to most (if not all) trainers in TRL.

> [!NOTE]
> [Since Transformers v5](https://github.com/huggingface/transformers/pull/42805), `from_pretrained` infers the dtype from the model's config (e.g., `bfloat16`) instead of
> defaulting to `float32`. When you load a model, pass `dtype` explicitly if you need a specific
> precision.
> When TRL handles model loading (i.e., you pass a model name string to the trainer), it defaults to
> `float32`.

## Use different optimizers and schedulers

By default, the [`DPOTrainer`] creates a `torch.optim.AdamW` optimizer. You can create and define a different optimizer and pass it to [`DPOTrainer`] as follows:
Expand Down