Pull Requests vllm-project/vllm

[Misc][Quantization] Clarify the intent of GGUF `FusedMoE` weight materialization ready

#30310 opened 2025-12-09 06:40 by a4lg

Fix incomplete response generation for tool call outputs frontend deepseek fb-exported meta-exported

#30304 opened 2025-12-09 05:01 by qandrew

[Misc] Pass reasoning to deepseekV32 tokenizer frontend deepseek

#30302 opened 2025-12-09 04:48 by kingsmad

[ResponsesAPI] Add GPTOSS MCP tool streaming frontend gpt-oss

#30301 opened 2025-12-09 04:36 by qandrew

[Bugfix] Update WSL detection to check for WSL1 compatibility as WSL2…

#30299 opened 2025-12-09 03:59 by HoneyBerries

[CI/Build][Kernel][BugFix][AMD] Fix per_token_group_quant_fp8 to use correct fp8 min/max values and update atol/rtol in test_quantfp8_group_functionality rocm

#30292 opened 2025-12-09 03:07 by rasmith

[CI/Build][AMD] Fix ref_dynamic_per_token_quant reference implementation on ROCm. rocm

#30291 opened 2025-12-09 02:53 by rasmith

[Core] Add token-level KV cache metrics to V1 engine v1

#30289 opened 2025-12-09 02:19 by Minsung-commit

Adding quantized fused_moe_lora support

#30286 opened 2025-12-09 01:22 by yugong333

Ensure minimum frames for GLM 4.6V compatibility ready

#30285 opened 2025-12-09 00:23 by gh-wf

[BugFix] Lazy tokenizer init in StructuredOutputManager to prevent GGUF semaphore leak structured-output v1

#30284 opened 2025-12-09 00:00 by kitaekatt

[Small] Add comment for `parallel_config` in `FusedMoEModularKernel`

#30282 opened 2025-12-08 22:32 by yewentao256

[CI/Build] Ignore data_parallel_size_local

#30281 opened 2025-12-08 22:05 by rjrock

Add `moe_align_block_size_no_permute` for small batch size with large num_expert needs-rebase

#30280 opened 2025-12-08 21:49 by RunkaiTao

[CPU][Bugfix] Fix CPU Profiler issue needs-rebase v1

#30278 opened 2025-12-08 21:31 by zhili03

[ROCM][CI] Fix AMD Examples Test Group documentation rocm ready ci/build

#30276 opened 2025-12-08 19:35 by Concurrensee

[NIXL] refine decoder side post process for heterogeneous BlockSize and kv_layout v1 kv-connector

#30275 opened 2025-12-08 19:34 by xuechendi

[AMD] Amd/deepseek aiter fusions rocm needs-rebase v1 deepseek

#30274 opened 2025-12-08 19:33 by k50112113

[Bugfix] Temporarily disable group quant rms norm fusion

#30273 opened 2025-12-08 18:34 by ElizaWszola

[CI/Build] Use spawn subprocess for ROCm documentation rocm

#30272 opened 2025-12-08 18:16 by rjrock

[ROCm][CI][Bugfix] Multi-Modal Model Support Fixes and Attention Backend Improvements rocm needs-rebase ci/build multi-modality qwen

#30270 opened 2025-12-08 17:00 by AndreasKaratzas

[Frontend] Fixes anthropic streaming message_start usage nesting frontend ready

#30266 opened 2025-12-08 15:58 by bbartels

Fix 500 /tokenize errors and 400 v1/chat/completions errors when using truncate_prompt_tokens and sending /tokenize and v1/chat/completions requests under high concurrency frontend

#30264 opened 2025-12-08 14:41 by Soufiane-Ra

Multiple Hybrid KV Cache Coordinator v1

#30263 opened 2025-12-08 14:09 by roikoren755

Support TP which is not divded for NVFP4 kernels (flashinfer-cutlass) by adding dynamic padding nvidia

#30260 opened 2025-12-08 13:25 by danielafrimi

[Feature]: OpenTelemetry Metrics Support v1

#30258 opened 2025-12-08 11:45 by mladjan-gadzic

[ROCm] Use aiter.topk_sigmoid in llama4 rocm needs-rebase llama

#30255 opened 2025-12-08 11:07 by tpopp

gptq marlin quantization support for fused moe with lora

#30254 opened 2025-12-08 10:33 by Bhanu068

[gpt-oss] Add model_identity to system message retrieval for harmony chat template frontend gpt-oss

#30247 opened 2025-12-08 08:43 by lyuwen

[Bugfix] Fix fusion for VL models

#30244 opened 2025-12-08 07:47 by ElizaWszola