-
Notifications
You must be signed in to change notification settings - Fork 79
Pull requests: jjang-ai/vmlx
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Fix Qwen hybrid prompt-cache reuse with native MTP
#223
opened Jul 1, 2026 by
jakimoai
Collaborator
Loading…
engine: fix MLA paged prefix-cache reconstruct for GLM-5.2
#218
opened Jun 25, 2026 by
Deviad
Loading…
fix(flash-moe): support Step-3.7-Flash (step3p7) expert offload
#210
opened Jun 17, 2026 by
st-adam
Loading…
MiniMax-M3 vMLX runtime + MSA dual-cache through all cache tiers
#205
opened Jun 13, 2026 by
jjang-ai
Owner
Loading…
WIP: live-app testing harness + protocol/status docs + VL stream-fix attempts
#204
opened Jun 13, 2026 by
jjang-ai
Owner
Loading…
Add Makefile dev install loop and panel build ergonomics
#195
opened Jun 10, 2026 by
unixwzrd
Loading…
5 tasks
Fix Qwen35 VL MTP runtime compat for hybrid SSM paths
#194
opened Jun 10, 2026 by
unixwzrd
Loading…
2 tasks
Fix Flash MoE MXFP8 expert matmul for JANG bundles
#193
opened Jun 10, 2026 by
unixwzrd
Loading…
2 tasks
feat(distributed): add VMLX_BACKEND env var for mlx.distributed backend
#172
opened May 23, 2026 by
kqb
Loading…
feat(pflash): importance-scored sparse prefill scaffold (#136)
#161
opened May 12, 2026 by
st-adam
Loading…
4 of 6 tasks
feat(pld): hybrid partial-accept replay for SSM models (#134)
#149
opened May 7, 2026 by
st-adam
Loading…
3 tasks
ProTip!
Adding no:label will show everything without a label.