Commits

History for

Commits on Jul 28, 2026

Support Qwen3-32B: fix split-K linear grid for non-dividing K dims (#744)
jyizheng
authored
adding inkling, GLM, optimized kimi dflash (#745)

olivia111
and
claude
authored

Commits on Jun 15, 2026

Commits on Jun 10, 2026

Commits on Jun 9, 2026

Feat/tensor view core (#711)

ZephyrZhao247
and
claude
authored

Commits on Jun 8, 2026

[MPK] Support EAGLE3 Speculative Decoding for Qwen3 (#692)

Risc-lt
and
claude
authored

Commits on Jun 4, 2026

fix: clamp TMA box dims for splitk linear on SM100 to fix deadlock (#705)
Electronic-Waste
authored

Commits on May 27, 2026

[Online Serving] Continuous batching, streaming, and OpenAI server support (#660)

authored

Commits on Apr 29, 2026

fix: recover test mode (#673)
ZephyrZhao247
authored

Commits on Apr 27, 2026

Add load_mpk for kernel reuse (#618)

AravTewari
and
jiazhihao
authored

Commits on Apr 24, 2026

DeepSeek V3 on B200: FP8 + MLA TP + MLA prefill + MTP (#667)

authored

Commits on Apr 19, 2026

fixes (#661)
Night-Traders-Dev
authored

Commits on Apr 17, 2026

Commits on Apr 14, 2026

cleanup (#655)
taras-sereda
authored

Commits on Apr 10, 2026

Add MLA multi-token decoding kernel for MTP verification (#651)
xinhaoc
authored

Commits on Apr 9, 2026

Commits on Mar 26, 2026

Dynamically Sharded Model Weight Loader (#632)

EH67
and
Emily Ho
authored

Commits on Mar 9, 2026

Add release wheel workflow (#637)
xinhaoc
authored

Commits on Feb 20, 2026

Commits on Feb 19, 2026

Fix: make non-editable pip install produce a working package, tests, build (#625)
Marco-Christiani
authored

Commits on Feb 13, 2026

mpk: lazily initialize CUDA driver in cuda_utils (#619)
ZephyrZhao247
authored

Commits on Feb 10, 2026

NVSHMEM tile-based allreduce (#614)
ZephyrZhao247
authored

Commits on Jan 29, 2026

Refactoring old allreduce tasks and improve performance (#612)
ZephyrZhao247
authored

Commits on Jan 25, 2026

fix: output on hopper is now correct (#610)
ZephyrZhao247
authored

Commits on Dec 20, 2025

Vllm integration -- Mirage side modification (#537)

authored

Commits on Dec 8, 2025

Replace rotary_embedding_sm100 with rotary_embedding_hopper && Update residual adding logic (#593)
yzhou442
authored

Commits on Nov 26, 2025

ADD split KV cache for Attention kernel (#579)

xinhaoc
and
yzhou442
authored

Commits on Nov 25, 2025

Support Gumbel-Max Sampling Kernel (#574)

STWMichael
and
Michael Wang
authored

Commits on Nov 18, 2025

Multi GPU fix (#580)
ZephyrZhao247
authored

Commits on Nov 11, 2025

[MPK] Qwen3-30B-A3B MoE demo (#563)

authored