Skip to content

Supported native Windows / PyTorch 2.14 GPT-OSS 20B NF4/QLoRA training tuple? #12044

Description

@drasumisrl-png

Which exact supported Unsloth and unsloth-zoo release or immutable revision supports GPT-OSS 20B NF4/QLoRA training on native Windows x64, Python 3.12, with PyTorch 2.14.0+cu130?

We found Unsloth PR #11016 (open) and Zoo PR #1227 (merged), but their reported checks do not establish GPT-OSS expert/attention backward on this target. Could you identify the supported dependency combination and an existing GPT-OSS-specific training recipe or validation?

We intend to train q_proj/v_proj LoRA r=4, alpha=8 on a 16 GiB GPU target; is that configuration covered, and what documented restrictions apply? We have not executed or validated this combination.

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions