Skip to content

Extend LLMFactory support to the post-training repo #432

New issue
New issue
Open
Open
Extend LLMFactory support to the post-training repo#432
Assignees
Neonkraft
Labels
4.6 post-training

Description

@Neonkraft
Neonkraft
opened on Oct 12, 2026

Goal
Support LlamaFactory in the post-training repo.

Description
Currently, the post-training repo primarily supports TRL. For training the flagship model, we want to support LlamaFactory as well.

Deliverable scope
Perform supervised fine-tuning (SFT) of Olmo-3 using LlamaFactory and achieve performance comparable to that of an equivalent checkpoint trained using TRL.

Dependencies
N/A

Activity

  1. Neonkraft
    self-assigned this
    on Oct 12, 2026
  2. Neonkraft
    added
    4.6 post-training
    on Oct 12, 2026
  3. Neonkraft
    added this to WP4 - LLM building task boardon Oct 12, 2026
  4. Neonkraft
    moved this to In Progress in WP4 - LLM building task boardon Oct 12, 2026

  5. Neonkraft commented on Oct 12, 2026

    @Neonkraft
    Neonkraft
    on Oct 12, 2026
    CollaboratorAuthor

    Work in progress in the Llamafactory branch

  6. Neonkraft commented on Oct 26, 2026

    @Neonkraft
    Neonkraft
    on Oct 26, 2026
    CollaboratorAuthor

    The PR for LlamaFactory can be found here. A trial run of tokenization for SFT using LlamaFactory was tested.

  7. zeppelinche-cmd
    moved this from In Progress to Review/Validation in WP4 - LLM building task boardon Oct 26, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

4.6 post-training

Type

No type

Projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions