Merge PoC_20260331 to sgl-project/sglang/main#282

Merged

Hexq0210 merged 43 commits intoAscend:release/PoC_20260415from

khalil2ji3mp6:main_merge_poc20260331

Apr 8, 2026

Commits on Mar 27, 2026

[NPU] Support GLM-4.7-Flash on NPU (#153)
Estrella-xx
authored

Commits on Mar 28, 2026

[NPU] recover accuracy for gemma3-4b-it from 54% to 72% (reduced by transformer5.3) (#155)
McZyWu
authored
Br fix qwen2 5 ascend (#159)
amote-i
authored
add documentation for GLM-4.7-Flash on Ascend (#162)
Estrella-xx
authored
Revert "Use LazyValue for routed_experts_weights_of_layer initialization" (#161)

longxin9715
and
j30065060
authored

Commits on Mar 30, 2026

fix(grok): fallback to standard weight loading when no presharded files found (#156)

Hide-on-bushsh
and
Jianzhao Xu
authored
revert: revert qwen3_5.py, use separate layers (#172)
iridiumine
authored
[bugfix]GLM-4V model (#176)

Hide-on-bushsh
and
Jianzhao Xu
authored
NPU can use piece cuda graph when the piece cuda graph is explicitly declared (#169)

chx96642264
and
chx96642264
authored
Bug fix for llama eagle3 (#177)
khalil2ji3mp6
authored
fix eagle3 accept rate (#179)
heziiop
authored
Support MTP for Qwen3.5 (#154)
iridiumine
authored
fix bug (#181)

longxin9715
and
j30065060
authored

Commits on Mar 31, 2026

Commits on Apr 1, 2026

Commits on Apr 2, 2026

[NPU] Use causal_conv1d and fix qwen-next modelslim (#207)
iridiumine
authored

Commits on Apr 3, 2026

Commits on Apr 6, 2026

adapt mtp + prefix for ascend gdn backend (#202)
silencejade
authored

Commits on Apr 7, 2026

Commits on Apr 8, 2026