weijinqian0 a6ef3ac4e4 [Performance] Pre-issued exponential distribution operator. (#4908) ...

Pre-issued exponential distribution operator.

Result:
Single inference saves 200-300 microseconds.
before：

<img width="2257" height="1058" alt="2"
src="https://github.com/user-attachments/assets/c1da19e2-a439-42cb-9d7c-c0218e61fd4c"
/>

After：

<img width="2211" height="342" alt="image"
src="https://github.com/user-attachments/assets/03c84292-c802-4755-949c-4266a9a72fc0"
/>


- vLLM version: v0.12.0
- vLLM main:
ad32e3e19c

---------

Signed-off-by: weijinqian_v1 <weijinqian@huawei.com>
Co-authored-by: weijinqian_v1 <weijinqian@huawei.com>

2025-12-11 23:02:51 +08:00

..

attention

[CI] speed up ut (#4901)

2025-12-11 18:45:43 +08:00

compilation

[UT]add pcp aclgraph ut (#4804)

2025-12-09 17:27:40 +08:00

core

Remove ascend schuduler ut (#4684)

2025-12-04 14:10:28 +08:00

device_allocator

add ut for device allocator/camem and mutistream/layers (#2037)

2025-07-31 19:17:27 +08:00

distributed

[Feat] Flashcomm2 use o_shared linear (#4188)

2025-12-11 12:43:04 +08:00

eplb

eplb redundant expert bugfix (#4291)

2025-11-21 14:24:35 +08:00

fake_weight

[CI] Add unit test framework (#1201)

2025-06-16 18:32:28 +08:00

kv_connector

[Bugfix] support mtp kv transfer and pp partition by hand in kv transfer (#4892)

2025-12-11 17:23:21 +08:00

model_loader/netloader

Drop 0.11.0 support (#4377)

2025-11-24 17:08:20 +08:00

ops

Drop torchair (#4814)

2025-12-10 09:20:40 +08:00

patch/worker/patch_common

[Refactor] refactor patch module (#3555)

2025-10-21 20:19:46 +08:00

quantization

[Feat] Support native Kimi-K2-Thinking native W4A16 quantized experts weights (#4516)

2025-12-10 15:58:52 +08:00

sample

[Performance] Pre-issued exponential distribution operator. (#4908)

2025-12-11 23:02:51 +08:00

spec_decode

upgrade vLLM to main (#4608)

2025-12-02 22:10:52 +08:00

worker

Remove mindie_turbo (#4896)

2025-12-11 18:46:12 +08:00

__init__.py

[2/4][Refactor] Refactor torchair utils (#1892)

2025-07-21 19:43:30 +08:00

base.py

[Feature]: implement the fusion of allreduce and matmul in prefill phase when tp is enabled (#1926)

2025-07-28 15:13:37 +08:00

conftest.py

[1/N][CustomOp] Register activation customop instead of overwrite forward_oot (#1841)

2025-07-18 23:07:14 +08:00

test_ascend_config.py

[Fusion] normalize fusion naming and enable e2e test (#4693)

2025-12-11 17:53:43 +08:00

test_envs.py

[Misc] Remove redundant imported envs, using envs_ascend instead (#2193)

2025-08-14 09:33:39 +08:00

test_platform.py

[bugfix][refactor] fix recompute_scheduler break with vllm 0.12.0 & support async scheduling & refactor recompute_scheduler.py (#4895)

2025-12-11 22:24:49 +08:00

test_utils.py

Remove mindie_turbo (#4896)

2025-12-11 18:46:12 +08:00