Andreas KaratzasandOpenAI Codex 394beb633b [Bugfix][ROCm] Use batch DMA for CPU KV cache loads (#49843) ...

Signed-off-by: Andreas Karatzas <akaratza@amd.com>
Co-authored-by: OpenAI Codex <noreply@openai.com>

2026-07-27 02:18:01 -07:00

..

assets

[Bugfix] Align OpenCV video metadata timeline (#47099)

2026-06-30 20:43:42 +00:00

benchmarks

[Bugfix][Benchmarks] Restore --skip-tokenizer-init with custom dataset (#49180)

2026-07-24 04:30:19 -07:00

compilation

[XPU] Enable QK Norm + RoPE fusion pass on XPU (#49394)

2026-07-27 03:22:29 +00:00

config

[Bugfix] Reject contradictory custom-op directives (#49134)

2026-07-26 08:42:14 -04:00

cute_utils

Revert "[Perf][GLM-5.2] Blackwell decode optimizations" (#49768)

2026-07-24 16:04:59 -07:00

device_allocator

[Bugfix][CuMem] Make KV-cache wake cleanup tag-safe (#49857)

2026-07-26 12:09:13 -07:00

distributed

[3/N][Core][KV Connector] Support reliable partial-tail KV offload for sub-block prompts (#49502)

2026-07-26 23:22:17 -07:00

engine

[Model] Add VaultGemma via Transformers modeling backend (#49803)

2026-07-25 16:54:15 +00:00

entrypoints

[Frontend] expose stream_interval as req sampling param (#49754)

2026-07-27 11:30:37 +08:00

inputs

[Render][Speculator] Add return_loss_mask to render endpoint for training data generation (#46846)

2026-06-28 00:07:33 -07:00

ir

[CI] Use explicit devices in IR tests (#49513)

2026-07-23 20:08:14 -07:00

kernels

[ROCm] Make vllm_c RMSNorm output contiguous (#49913)

2026-07-26 23:59:14 -05:00

logging_utils

[vLLM IR] 1/N Implement IR skeleton and rms_norm op (#33825)

2026-03-31 22:15:05 -04:00

lora

[Model] Add Inkling LoRA support [4/N] (#48884)

2026-07-17 09:52:15 +08:00

model_executor

[Bugfix][CPU] Zero-pad MoE intermediate size for grouped-gemm TP alignment (#49591)

2026-07-27 16:32:23 +08:00

models

[Perf] Fix moe reduce_scatter perf regression by removing additional comm, 5% E2E throughput gain back. (#48763)

2026-07-25 16:36:19 +00:00

multimodal

[multimodal] Make PyNvVideoCodec decoder concurrency configurable (#49753)

2026-07-24 23:31:41 -07:00

parser

[Bugfix][Parser] Fix special tokens (EOS/BOS) leaking into reasoning content (#48748)

2026-07-22 16:10:55 -04:00

platforms

[XPU] Enable QK Norm + RoPE fusion pass on XPU (#49394)

2026-07-27 03:22:29 +00:00

plugins

[Frontend] Add endpoint plugins framework (#47454)

2026-07-08 10:00:41 +08:00

profiler

[Refactor] Clean up log once scope="local" (#40540)

2026-04-22 16:42:43 -04:00

ray

Propagate Flash Attention cache configuration to Ray workers (#49177)

2026-07-21 07:47:53 +00:00

reasoning

[Model] Add Inkling model support [1/N] (#48799)

2026-07-15 23:40:07 -07:00

renderers

[Perf] Isolate MM preprocessing on its own executor (#49524)

2026-07-26 08:04:17 +00:00

third_party

[Refactor] Move fla to third party (#48500)

2026-07-16 19:22:36 +01:00

tokenizers

[Tokenizer] Use HF config for HF tokenizers (#49907)

2026-07-27 07:49:11 +00:00

tool_parsers

[Bugfix][Tool Parser] Fix dropped streaming arguments in Jamba and InternLM2 parsers (#48852)

2026-07-25 09:34:43 -04:00

tracing

…

transformers_utils

[Model] Add VaultGemma via Transformers modeling backend (#49803)

2026-07-25 16:54:15 +00:00

triton_utils

[Kernel] TD operand loads for batched MoE GEMM (moe_mmk) on XPU (#46340)

2026-07-26 08:50:07 +08:00

usage

replace cuda_device_count_stateless() to current_platform.device_count() (#37841)

2026-03-31 22:32:54 +08:00

utils

[WideEP] Update NCCL to 2.30.7 to enable DeepEPv2 in the vllm/vllm-openai image (#45321)

2026-07-24 13:00:02 -07:00

v1

[Bugfix][ROCm] Use batch DMA for CPU KV cache loads (#49843)

2026-07-27 02:18:01 -07:00

vllm_flash_attn

[CI/Build][Docker] Bump nvidia-cutlass-dsl to 4.6.0 and drop packaging workarounds (#47442)

2026-07-16 13:51:41 +00:00

__init__.py

…

_aiter_ops.py

[ROCM] Fix AITER Fused AllReduce RMSNorm for Transformers Backend (#49673)

2026-07-24 07:16:17 -07:00

_custom_ops.py

[CPU][Spec Decode] Optimize GDN conv path for speculative decoding (#48577)

2026-07-27 06:20:14 +00:00

_xpu_ops.py

[XPU][DeepSeekV4]Add DeepSeek-V4 fuse_index_q SYCL kernel path (#45991)

2026-07-21 21:22:34 +08:00

collect_env.py

[XPU]Enhance environment collection for Intel XPU and optimize layout (#35698)

2026-04-13 12:51:46 +08:00

connections.py

[CI] Add retry with 4x backoff to HTTP fetches for transient failures (#37218)

2026-03-19 19:00:20 +00:00

env_override.py

[Perf] Add VLLM_TRITON_FORCE_FIRST_CONFIG to skip Triton autotuning (#42425)

2026-06-16 15:16:45 +02:00

envs.py

[Bugfix] Exclude location-derived path vars from torch.compile cache factors (#47573)

2026-07-22 16:56:09 -07:00

exceptions.py

[Bugfix] Return HTTP 422 for unprocessable image URLs instead of 500 (#47165)

2026-07-06 11:56:23 +00:00

forward_context.py

[Feature] Support sequence parallel without the need for DP, 1.9%~5.0% E2E Throughput Improvement (#47070)

2026-07-05 05:41:30 -07:00

logger.py

[Log] Log once in local node by default (#37568)

2026-03-19 12:04:59 -07:00

logits_process.py

…

logprobs.py

…

model_inspection.py

…

outputs.py

[Front-end] [Messages] Populate num_cache_creation_tokens (#48535)

2026-07-18 13:04:35 -04:00

pooling_params.py

[Bugfix] Reject removed pooling parameters (#48984)

2026-07-19 05:18:03 -07:00

py.typed

…

sampling_params.py

[Frontend] expose stream_interval as req sampling param (#49754)

2026-07-27 11:30:37 +08:00

scalar_type.py

Humming support for 2/3/5/6/7-bit pack-quantized weight-only inference (#46389)

2026-06-24 13:53:54 -04:00

scripts.py

…

sequence.py

[Core] Cleanup KVConnector handling with PP + fix MRV2 (#43732)

2026-05-28 13:12:03 -07:00

tasks.py

[Bugfix] Reject removed pooling parameters (#48984)

2026-07-19 05:18:03 -07:00

version.py

…