forked from sgl-project/sglang
-
Notifications
You must be signed in to change notification settings - Fork 10
Pull requests: Ascend/sglang
Author
Label
Projects
Milestones
Reviews
Assignee
Sort
Pull requests list
qwen3.5-27b in256 testcase
ascend-cla/no
npu
#460
opened Apr 30, 2026 by
liuxianglong17
Loading…
5 tasks
Add GDN chunked prefill metadata management and refactor handling
ascend-cla/yes
npu
#443
opened Apr 29, 2026 by
AndyLi429
Loading…
5 tasks
use aisbench in deepep test cases
ascend-cla/no
deepseek
npu
#439
opened Apr 25, 2026 by
liuxianglong17
Loading…
5 tasks
Feat/ascend gdn chunk meta precompute
ascend-cla/yes
npu
piecewise-cuda-graph
#437
opened Apr 25, 2026 by
AndyLi429
Loading…
5 tasks done
[NPU] Support shared expert dual stream optimization and fix DP Attention compatibility
ascend-cla/no
#434
opened Apr 24, 2026 by
iridiumine
Loading…
3 of 5 tasks
add num_shot=5 in mmlu test for deepseek v2
ascend-cla/no
deepseek
npu
#430
opened Apr 23, 2026 by
liuxianglong17
Loading…
5 tasks
[test] Pr to ascend 0406
ascend-cla/no
deepseek
lora
npu
#422
opened Apr 22, 2026 by
litmei
Loading…
5 tasks
use triton split_qkvgate_gemma_rmsnorm_rope
ascend-cla/yes
#412
opened Apr 21, 2026 by
Liwansi
Loading…
5 tasks
Bugfix: Qwen3-VL-MoE adapt encoder_only
amd
ascend-cla/no
blackwell
deepseek
dependencies
diffusion
documentation
Improvements or additions to documentation
jit-kernel
Multi-modal
npu
piecewise-cuda-graph
quant
sgl-kernel
speculative-decoding
#407
opened Apr 20, 2026 by
Hide-on-bushsh
Loading…
5 tasks
Add barriers for shared memory in DP and non-DP attention
ascend-cla/no
#394
opened Apr 17, 2026 by
AndyLi429
Loading…
5 tasks
[NPU] fix global expert distribution recording num_tokens_per_rdma_rank error in normal DeepEP mode
ascend-cla/no
#385
opened Apr 15, 2026 by
ZeyuanChen2000
Loading…
5 tasks
Fix the parsing error for Qwen3.5 thinking models.
ascend-cla/no
#352
opened Apr 11, 2026 by
iridiumine
Loading…
2 of 5 tasks
support trinity-mini model for npu
ascend-cla/no
npu
#310
opened Apr 9, 2026 by
McZyWu
Loading…
5 tasks
solve remaining accuracy problem introduced by loading weights for pr 19321
amd
ascend-cla/no
blackwell
deepseek
dependencies
diffusion
documentation
Improvements or additions to documentation
hicache
jit-kernel
lora
model-gateway
Multi-modal
npu
quant
sgl-kernel
#248
opened Apr 7, 2026 by
McZyWu
Loading…
5 tasks
Add npu performance testcases and workflow.
ascend-cla/no
deepseek
lora
npu
#232
opened Apr 6, 2026 by
shun8686
Loading…
5 tasks
[model] support trinity-mini for npu accuracy 90%
ascend-cla/no
npu
#206
opened Apr 2, 2026 by
McZyWu
Loading…
5 tasks
Revert "Revert "Use LazyValue for routed_experts_weights_of_layer initialization""
ascend-cla/no
#189
opened Mar 31, 2026 by
Hexq0210
Loading…
[NPU] fix VoxtralRealtimeTextModel miss in 5.3.0 transformers version for model Mistral-Small-3.1-2506
ascend-cla/no
npu
#175
opened Mar 30, 2026 by
ZeyuanChen2000
Loading…
5 tasks
Previous Next
ProTip!
Find all pull requests that aren't related to any open issues with -linked:issue.