-
-
Notifications
You must be signed in to change notification settings - Fork 123
Pull requests: turboderp-org/exllamav3
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
Fix MTP speculative decoding on multi-GPU splits (fixes #260)
#262
opened Jul 28, 2026 by
dushyant30suthar
Loading…
Allow using multiple draft models at once
#261
opened Jul 28, 2026 by
Ednaordinary
•
Draft
2 of 5 tasks
Add Gemma4 E4B and E2B support (per-layer embeddings, cross-layer KV sharing)
#253
opened Jul 25, 2026 by
pdcruze
Loading…
Cache/Attn: Load Q8 cache bytes directly
#247
opened Jul 18, 2026 by
remichu-ai
Contributor
Loading…
[kernel] Persistent route-packed (route_block x n_tile) schedule for exl3 fused MoE (sm120)
#246
opened Jul 15, 2026 by
brandonmmusic-max
Loading…
Detailed size estimation + fix #193 (overestimating fused QKV storage)
#194
opened Apr 16, 2026 by
mratsim
Contributor
Loading…
ProTip!
Exclude everything labeled
bug with -label:bug.