{{ message }}
-
Notifications
You must be signed in to change notification settings - Fork 53
Pull requests: mudler/vllm.cpp
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
feat(KERNEL-QUANT-CIQ-GEMM-ROCM-IQUANT): port IQ4_XS/IQ3_XXS to ROCm
#3029
opened Sep 6, 2026 by
joral
Contributor
Loading…
feat(BACKEND-TENSTORRENT-KEEPQUANT): Q5_K/Q6_K/Q8_0 decode chains and the q4km vehicle gate (#2959)
#3028
opened Sep 6, 2026 by
lu-zero
Collaborator
Loading…
perf(BACKEND-ROCM): widen sampling block to 1024 and split-phase random sample
#3010
opened Sep 6, 2026 by
ghazni101
Contributor
Loading…
fix(ORACLE-PIN-SOURCE-BUILD): replace the stale pin inequality with the guarantee it stood for, and unred main
#2941
opened Sep 5, 2026 by
localai-org-maint-bot
Collaborator
Loading…
feat(BACKEND-ROCM): verify Qwen capture provenance
#2932
opened Sep 4, 2026 by
VikashLoomba
Contributor
•
Draft
fix(MODEL-TEXT-GLM4-MOE-LITE-GATE-2839): the near-tie predicate cannot fire, so apply the bar the oracle capture licenses -- and it FAILS 69/128
#2906
opened Sep 4, 2026 by
localai-org-maint-bot
Collaborator
Loading…
fix(GFX1100-TG200): skinny_gemm OOB write + MoE silu gate rounding
#2894
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
record(GFX1100-TG200): GDN b/a GEMV merge closed negative (VT_GDN_MERGED_BA_ROCM)
#2893
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
feat(GFX1100-TG200): fuse silu-mul with Q8_K quant epilogue (VT_SILU_QUANT_FUSED)
#2890
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T14 row-split greedy argmax arm
#2876
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T9 cooperative gated norm arm
#2875
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T8 cooperative single-row rmsnorm arm
#2874
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T6b cooperative attn preamble arm
#2868
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T6a cooperative GDN scan arm
#2866
opened Sep 4, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T5a vectorizes shared Q8_K quant superblock body
#2822
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
feat(GFX1100-TG200): T25 keep ssm_out as Q5_K with runtime input permutation
#2807
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
feat(GFX1100-TG200): T21 keep-quant for V-head row-permuted GDN projections
#2804
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
feat(GFX1100-TG200): T24 LDS-buffered quant epilogue in RmsNormRowCoopKernel
#2800
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T27 warp-cooperative QuantizeQ8KK for decode
#2796
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
feat(GFX1100-TG200): lever-C fused norm-quant epilogue
#2792
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T4a MMVQ K-quant decode GEMV arm
#2790
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T16 YTILE=4 default for wvSplitK decode-skinny GEMV
#2787
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
feat(KERNEL-QUANT-CIQ-GEMM-ROCM): land the W1 keep-quant providers on kROCM
#2782
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
Port async device-mirror combine/scatter kernels to ROCm
#2779
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
perf(GFX1100-TG200): T2b flips ROCm support_static_graph_mode
#2777
opened Sep 3, 2026 by
ghazni101
Contributor
Loading…
Previous Next
ProTip!
Follow long discussions with comments:>50.
You can’t perform that action at this time.
