Skip to content

Add instructions for the post-training steps#222

Merged
cmpatino merged 3 commits intomainfrom
post-training-instructions
Jul 29, 2025
Merged

Add instructions for the post-training steps#222
cmpatino merged 3 commits intomainfrom
post-training-instructions

Conversation

@cmpatino
Copy link
Contributor

No description provided.

@cmpatino cmpatino requested review from lewtun and loubnabnl July 24, 2025 21:39
Copy link
Member

@lewtun lewtun left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM with a sanity check that the mid-train and SFT configs yield a GBS of 128 on 1 node (feel free to merge once that's checked)


```shell
# Step 1 - Mid-Training
ACCELERATE_LOG_LEVEL=info accelerate launch --config_file recipes/accelerate_configs/zero3.yaml scripts/sft.py --config recipes/smollm3/sft/mid.yaml
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just to check: I ran the mid-training on 8 nodes, so here we should override the GAS to ensure we have a GBS of 128

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

That was a good spot! Checked and changed for mid-training, SFT, and DPO commands.

@cmpatino cmpatino merged commit 925acdf into main Jul 29, 2025
2 checks passed
@cmpatino cmpatino deleted the post-training-instructions branch July 29, 2025 10:10
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants