vllm-project/vllm

Pull Requests Commits

Revert "[Not to Main] Docker Release Mooncake Deps (#159)"

ywang96 committed 79 days ago

71475ffb

Sanitize unfilled recv slots in flashinfer_nvlink_one_sided dispatch (#9)

liuzijing2014 committed 79 days ago

17bcc8ee

mxfp8 dispatch support

zyongye committed 79 days ago

607f3a50

Add bf16 + defer-input-quant support to flashinfer_nvlink_one_sided all2all

zyongye committed 79 days ago

9167ef8d

[Not to Main] Docker Release Mooncake Deps (#159)

zhewenl committed 79 days ago

d7045619

[KV Connector] Pre-fill DecodeBenchConnector KV cache once at registration

liuzijing2014 committed 79 days ago

206aee7d

[KV Connector] Opt DecodeBenchConnector into SupportsHMA

liuzijing2014 committed 79 days ago

674e6ffd

[Bugfix] Fix max_num_batched_token not captured in cuda graph (#40734)

wzhao18 committed 79 days ago

Verified 8b49cf3a

[Bugfix] fix inductor error for dpsk v4 (#41135)

ZJY0516 committed 79 days ago

Verified 2ae73c75

[BugFix][CPU] fix error on CPU runner shutdown (#41034)

fadara01 committed 79 days ago

Verified d95d03c7

[Bugfix] Fix Deepseek V4 import error due to AOT compile cache loading (#41090)

wzhao18 committed 79 days ago

Verified 803b9d78

[Feature] add cohere reasoning and tool parsers (#40422)

walterbm committed 79 days ago

Verified 1312f075

[BE][Torch 2.12] Remove workaround code for fixed cublas issue (#40845)

Lucaskabela committed 79 days ago

Verified fa1b9840

[QeRL] Add warnings for extra memory buffering (#40309)

kylesayrs committed 79 days ago

Verified 916e56c0

[Docs] [QeRL] Layerwise Reloading Documentation (#40317)

kylesayrs committed 79 days ago

Verified a085b525

uncomment flex backend for batch invariant mode (#40842)

liangel-02 committed 79 days ago

Verified 7fd05e05

[UX] Allow enable/disable model weights loading tracking by config (#41086)

Isotr0py committed 79 days ago

Verified 99255f3c

[CI] De-flake test_chat_completion_n_parameter_non_streaming (#41147)

haosdent committed 79 days ago

Verified 75a7cf2c

[CI] Return HTTP 400 for unsupported chat content part type (#41121)

haosdent committed 79 days ago

Verified 4b95e9ce

[CI][AMD][BugFix] Patch has_flashinfer decorator for test_select_rocm_aiter_backend (#41072)

rasmith committed 79 days ago

Verified 856b15c6

[DSV4] Align aux stream API with DeepseekV4DecoderLayer (#41171)

zixi-qi committed 79 days ago

Verified 6fb3f7b4

[Bugfix][ROCm] Fix gemm_a4w4 call to use updated AITER API signature (#40754)

chelnnexy committed 79 days ago

Verified d109eacd

[Core] Account for `num_gpu_blocks_override` in `max_model_len` checks (#41069)

njhill committed 79 days ago

Verified e68fa1b9

[Doc] Add missing API endpoints to security documentation (#40532)

russellb committed 79 days ago

Verified f05f3664

[FEATURE] Add EagleMistralForCausalLM (#41024)

juliendenize committed 79 days ago

Verified e9f8f31e

[Bugfix] release KV blocks for skipped P-ranks to prevent invalid KV errors and timeouts when P_tp > D_tp and MLA (#40449)

yangrz7 committed 79 days ago

Verified de3fe8dc

[New Model] Laguna XS.2 implementation (#41129)

joerowell committed 79 days ago

Verified 0899f436

[CI][AMD][BugFix] Update request URL in test_moriio_connector to match vllm-router compatibility changes (#41076)

rasmith committed 79 days ago

Verified 358a755e

[Build] Defer flashinfer cubin download to avoid ~2.5 GB (decompressed) layer duplication (#41134)

benoittgt committed 79 days ago

Verified a6088364

[DSV4] Enable Multi-stream for Pre-Attn GEMM (#41061)

zyongye committed 79 days ago

Verified 5aa371dc

Older