Skip to content

Fix PPO example #4556

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

Merged
kashif merged 4 commits into from
Nov 26, 2025
Merged

Fix PPO example #4556

Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
48ab047
Fix PPO example
qgallouedec Nov 21, 2025
19d7bf2
fix example
qgallouedec Nov 21, 2025
ad7db84
Merge branch 'main' into fix-ppo-example
qgallouedec Nov 24, 2025
271d1cc
Merge branch 'main' into fix-ppo-example
qgallouedec Nov 25, 2025
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
10 changes: 3 additions & 7 deletions examples/scripts/ppo/ppo_tldr.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,7 +44,7 @@

"""
python examples/scripts/ppo/ppo_tldr.py \
--dataset_name trl-internal-testing/tldr-preference-sft-trl-style \
--dataset_name trl-lib/tldr \
--dataset_test_split validation \
--learning_rate 3e-6 \
--output_dir pythia-1b-deduped-tldr-preference-sft-trl-style-ppo \
Expand All @@ -62,7 +62,7 @@

accelerate launch --config_file examples/accelerate_configs/deepspeed_zero2.yaml \
examples/scripts/ppo/ppo_tldr.py \
--dataset_name trl-internal-testing/tldr-preference-sft-trl-style \
--dataset_name trl-lib/tldr \
--dataset_test_split validation \
--output_dir pythia-1b-deduped-tldr-preference-sft-trl-style-ppo \
--learning_rate 3e-6 \
Expand Down Expand Up @@ -134,11 +134,7 @@ def prepare_dataset(dataset, tokenizer):
"""pre-tokenize the dataset before training; only collate during training"""

def tokenize(element):
input_ids = tokenizer.apply_chat_template(
element["messages"][:1],
padding=False,
add_generation_prompt=True,
)
input_ids = tokenizer(element["prompt"], padding=False)["input_ids"]
return {"input_ids": input_ids, "lengths": len(input_ids)}

return dataset.map(
Expand Down
Loading