-
Notifications
You must be signed in to change notification settings - Fork 113
Pull requests: youssofal/MTPLX
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
request_capture: persist exact completion token ids
#330
opened Aug 22, 2026 by
CharliePetch
Loading…
gdn_capture: fused_add_rmsnorm(threadgroup_size=512) perturbs prefill above 64 rows
#319
opened Aug 21, 2026 by
grzracz
Loading…
graphbank: batch the paged-offset device reads in _resolve_bucket
#318
opened Aug 21, 2026 by
grzracz
Loading…
Opt-in load-time GDN projection fusion on the verify-kernel path
#316
opened Aug 21, 2026 by
grzracz
Loading…
Speculative accept: batch the greedy argmax host round trips
#315
opened Aug 21, 2026 by
grzracz
Loading…
profiles: sustained/turbo set a batched target-distribution lane they also disable; allow operator override
#314
opened Aug 21, 2026 by
grzracz
Loading…
Speculative draft: run the greedy MTP chain on device, one host sync per cycle
#313
opened Aug 21, 2026 by
grzracz
Loading…
Qwen3.8 DFlash2 benchmark against Optimized Speed and Quality MTP
#304
opened Aug 20, 2026 by
davidtai
Contributor
Loading…
Make the 6-bit hexpack route at m == 4 opt-in: measured as a loss on a real 27B trunk
#302
opened Aug 20, 2026 by
Blakeolson21
Loading…
Hold the background warmup ladder while a request is in flight
#300
opened Aug 20, 2026 by
Blakeolson21
Loading…
fix(smart-fan): stop a wedged request pinning fans at max
#295
opened Aug 19, 2026 by
El-Patronum
Loading…
perf(mtp): combine greedy token and confidence reads (+0.65% throughput)
#288
opened Aug 18, 2026 by
ArthurOstapenko
Loading…
Fix Gemma 4 session cache across rewritten conversation turns
#283
opened Aug 18, 2026 by
TipKnuckle
Loading…
fix(forge): honor quantize:false module overrides
#281
opened Aug 17, 2026 by
shiftedx
Contributor
Loading…
feat(bench): add --compare-static fixed-depth baselines to mtp-adaptive
#276
opened Aug 17, 2026 by
rinaldofesta
Loading…
fix(server): preserve native tools for parallel requests
#270
opened Aug 17, 2026 by
shiftedx
Contributor
Loading…
build(deps): update transformers requirement from !=5.13.0,<5.15 to !=5.13.0,<5.16
dependencies
Pull requests that update a dependency file
python
Pull requests that update python code
#260
opened Aug 15, 2026 by
dependabot
Bot
Loading…
Add DeepSeek V4 target-only mlx-serve backend
#254
opened Aug 14, 2026 by
PhilipJohnBasile
Contributor
Loading…
Integrate DeepSeek V4 Flash 0731 into MTPLX
#251
opened Aug 13, 2026 by
davidtai
Contributor
Loading…
Previous Next
ProTip!
Exclude everything labeled
bug with -label:bug.