Qwen3 missing optimizations #1453

New issue

Copy link

New issue

Copy link

Open

Open

Qwen3 missing optimizations#1453

Copy link

Description

avik-pal

opened on Aug 30, 2025

Issue body actions

• faster SDPA from LuxLib feat: faster scaled_dot_product_attention for reactant #1452
• compile the generation loop instead of single token compile
• implement KV caching

👍React with 👍1yeruoforever

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