vllm-project/vllm

Pull Requests Commits

single deepep handle

LucasWilkinson committed 242 days ago

de92ab52

Revert "Merge commit '6e8d8c4afbddf725b34ef938616701869f5b3462' into sage/dbo-full-cudagraphsh"

yewentao256 committed 242 days ago

9f04a6cf

Revert "fix ubatch datatype issue"

yewentao256 committed 242 days ago

a0a11bc0

fix full cudagraphs for cutlass mla

SageMoore committed 248 days ago

143b09e6

add support for cutlass mla full cudagraphs

SageMoore committed 248 days ago

090f485a

revert kv connector fix

SageMoore committed 248 days ago

6d76bd03

fix ubatch datatype issue

yewentao256 committed 248 days ago

9e16220e

Merge commit '6e8d8c4afbddf725b34ef938616701869f5b3462' into sage/dbo-full-cudagraphsh

yewentao256 committed 248 days ago

5215c80a

fix assert error num_tokens_across_dp is None

yewentao256 committed 250 days ago

dd2a94fd

fix num_tokens_across_dp sizing issue

SageMoore committed 250 days ago

e526b1c0

fix set forward context error

yewentao256 committed 250 days ago

44ead56a

Fix pre-commit error

yewentao256 committed 250 days ago

28e7c30b

remove debug logging

SageMoore committed 253 days ago

2cf200c5

add support for multiple builders in the model runner

SageMoore committed 253 days ago

5bbfd95b

SageMoore committed 256 days ago

6b0c303a

SageMoore committed 256 days ago

4819bb87

[Attention][DBO] Add support for "splitting" the CommonAttentionMetadata (#21153)

SageMoore committed 260 days ago

Verified 0edaf752

[Test] Add Unit Test for Batched DeepGEMM (#21559)

yewentao256 committed 260 days ago

Verified 6e8d8c4a

[BugFix] Improve internal DP load balancing (#21617)

njhill committed 260 days ago

Verified 8d524ce7

[Speculators][Speculative Decoding] Add Qwen Eagle3 Support (#21835)

dsikka committed 260 days ago

Verified 9f9c38c3

[Misc] DeepGemmExperts : Avoid JIT generation in the hot-path (#21955)

varun-sundar-rabindranath committed 260 days ago

Verified a65f46be

[Misc] `VLLM_TARGET_DEVICE.lower()` (#22101)

NickLucche committed 260 days ago

Verified 57393715

[Model] Qwen2.5 VL SiLU-and-Mul (#22066)

vllmellm committed 260 days ago

Verified ee2eb6ec

[V1][CUDA] Full cudagraph support for FlashInfer (#21367)

fhl2000 committed 260 days ago

Verified 23322431

feat: Add Support GPTQ Quantization MOE on ROCM vllm serve (#21733)

JartX committed 260 days ago

Verified 3654847d

[Perf] Optimize `reshape_and_cache_flash` CUDA Kernel (#22036)

yewentao256 committed 260 days ago

Verified eefbf4a6

[CI] Initial tests for SM100 Blackwell runner (#21877)

mgoin committed 260 days ago

Verified 88faa466

[BugFix] Harden distributed DP startup (#21538)

njhill committed 260 days ago

Verified 881e1af4

Add lora test for tp>1 case for TPU. (#21970)

vanbasten23 committed 260 days ago

Verified d84b97a3

Introduce RayPPCommunicator for ray-based PP (#21660)

ruisearch42 committed 260 days ago

Verified d3317594

Older