Repository navigation

Gemma4 support #1062

Closed

Assignees

Labels

Milestone

opened

Export gemma4 E2B and E4B models. Key to this is efficient memory management, as a huge part of model params (above 50%) is not needed at runtime, but only on first token consumption (PLE layers)

mkopcins
self-assigned this
on Apr 8, 2026
msluszniak
added
modelIssues related to exporting, improving, fixing ML models
on Apr 8, 2026
msluszniak
added the
Feature
issue type on Apr 8, 2026
ninjapayofficial commented on Apr 11, 2026
ninjapayofficial
on Apr 11, 2026
yeah Gemma is needed
Reacted by Mateusz Sluszniak, sanjiv kumar pandit and Alexey Balmasov
Andrewgrillo23 commented on Apr 14, 2026
Andrewgrillo23
on Apr 14, 2026
Truly excited for this one!
Reacted by Mateusz Sluszniak and sanjiv kumar pandit
msluszniak
linked a pull request that will close this issuefeat: Adding support for gemma4 e2b models #1162 on Jun 1, 2026
msluszniak
added this to the v0.10.0 milestone on Jun 9, 2026
mkopcins
closed this as completedin #1162 on Jun 11, 2026
mkopcins
added a commit that references this issue on Jun 11, 2026
feat: Adding support for gemma4 e2b models (#1162)
7d145b5
msluszniak
modified the milestones: v0.10.0, v0.9.0 on Jul 23, 2026

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

Metadata

Assignees

mkopcins

Labels

Type

Projects

No projects

Milestone

v0.9.0

Relationships

None yet

Development

No branches or pull requests