[Model] [Config] Correctly identify granite-4.0-micro as non-hybrid model (#28563)

Thomas Parnell committed about 1 year ago

[EPLB][ROCm]: support EPBL for ROCm backend (#27731)

PerryZhang01 committed about 1 year ago

Implement ARC KV cache eviction policy for CPU offloader (#27039)

alberto committed about 1 year ago

[Perf] Refactor cudagraph_support to enable full CUDA graphs for spec decoding with FlashInfer (#28479)

Benjamin Chislett committed about 1 year ago

Remove deprecated fields from `CompilationConfig` (#27593)

Harry Mellor committed about 1 year ago

[Misc] Refactor Attention kv transfer methods into decorator (#27816)

Nicolò Lucchesi committed about 1 year ago

[Refactor] Remove redundant TP gather/split in split_qkv in QwenVL (#28271)

Canlin Guo committed about 1 year ago

[Performance][Hopper] Avoid M dim padding to 4x for most cases (due to cuda graphs paddings) (#28492)

Alexander Matveev committed about 1 year ago

Fix pre-commit (and XPU) on `main` (#28556)

Harry Mellor committed about 1 year ago

[V0 deprecation] Deprecate use_v1 parameter (#28112)

wangxiyuan committed about 1 year ago

[Bugfix] Fix gpt_oss packed_modules_mapping (#28536)

Jee Jee Li committed about 1 year ago

[ROCm] [Bugfix] Fix `fused_qknorm_rope_kernel` rocm compatibility (#28500)

TJian committed about 1 year ago

add cpu option for p/d in nixl_connector (#28356)

ZhengHongming888 committed about 1 year ago

[KVConnector] Enable get_block_ids_with_load_errors() in LMCache connector (#27978)

ziruiliu committed about 1 year ago

[XPU]Fix crash due to removed VLLM_USE_V1 attribute (#28520)

Chaojun Zhang committed about 1 year ago

[Model] fix glm4_moe_mtp load weights with GLM-4.6 checkpoint. (#27597)

wuyaoxuehun committed about 1 year ago

[BugFix]: --enable-lora with model granite-4.0-micro crash (#27733)

yyzxw committed about 1 year ago

[CI] Introduce autorun_on_main feature (#27836)

Huamin Li committed about 1 year ago

[CI/Build] Fix crash due to removed VLLM_USE_V1 attribute in EPD (#28521)

Chenguang Zheng committed about 1 year ago

[Core] Cache `vllm_is_batch_invariant` (#28304)

Lukas Geiger committed about 1 year ago