stefankoncarevicandAndreas Karatzas b369f10d5c [Bugfix][ROCm][CI] Restore the DeepSeek-V4 input GEMM override point (#51821) ...

Signed-off-by: Stefan Koncarevic <Stefan.Koncarevic@amd.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>

2026-08-12 20:33:52 -05:00

..

assets

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

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

benchmarks

[Profiler] Add minimal Triton Proton profiling backend (#48789)

2026-08-11 14:51:57 -07:00

compilation

[ROCm] Fix AITER all-reduce fusion coverage (#50802)

2026-08-06 14:52:23 +08:00

config

[Spec Decode] DSpark confidence-scheduled verification (#47808)

2026-08-12 12:27:48 -07:00

cute_utils

[Kernel][SM100] Add a CuTeDSL fused query kernel (#49792)

2026-08-04 15:23:06 -07:00

device_allocator

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

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

distributed

[Docs][RL] Rewrite weight-transfer docs; standardize examples (#51729)

2026-08-11 18:53:15 -07:00

engine

[Config] Update default _max_num_batched_tokens from 8192 to 16384 (#51726)

2026-08-11 11:09:34 -04:00

entrypoints

[Bugfix] Bound Anthropic stop sequences (#51997)

2026-08-12 12:05:12 -04:00

inputs

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665)

2026-07-29 04:20:40 +00:00

ir

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

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

kernels

[Bugfix][Triton] Make fp8_min/fp8_max constexpr in _quantize_pad_fp8_kernel (#51872)

2026-08-12 03:09:54 +00:00

logging_utils

…

lora

[Refactor] Delete dead code in models (#51838)

2026-08-11 15:44:26 -04:00

model_executor

[Spec Decode] DSpark confidence-scheduled verification (#47808)

2026-08-12 12:27:48 -07:00

models

[Bugfix][ROCm][CI] Restore the DeepSeek-V4 input GEMM override point (#51821)

2026-08-12 20:33:52 -05:00

multimodal

[Bugfix][Multimodal] Invalidate retained PyNvVideoCodec decoder after failure (#51139)

2026-08-12 11:02:10 +00:00

parser

[Bugfix] Align deepseek v4 parser thinking default with tokenizer (#51296)

2026-08-10 19:30:30 +00:00

platforms

[ROCm] Enable DeepSeek-V4 on gfx11 (#47017)

2026-08-12 08:55:10 +08:00

plugins

[Frontend] Add endpoint plugins framework (#47454)

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

profiler

[Profiler] Add minimal Triton Proton profiling backend (#48789)

2026-08-11 14:51:57 -07:00

ray

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

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

reasoning

[Model][Frontend] Add Ling 3.0 Flash BF16, MTP, and parser support (#51045)

2026-08-06 00:38:28 +08:00

renderers

Preserve revision pins in secondary artifact loaders (#51446)

2026-08-11 04:33:10 +00:00

third_party

[New model] Kimi K3 (#50000)

2026-07-30 03:49:58 -07:00

tokenizers

Bound generation inputs before expensive work (#51447)

2026-08-11 22:03:50 +08:00

tool_parsers

[Model] Add native Dots3 NOTE multimodal support (#51255)

2026-08-12 23:13:20 +08:00

tracing

…

transformers_utils

[Model] Add native Dots3 NOTE multimodal support (#51255)

2026-08-12 23:13:20 +08:00

triton_utils

[ROCm] [Model] Enable TML inkling (#48841)

2026-07-27 10:05:46 +00:00

usage

[Bugfix] Close usage telemetry HTTP sessions (#51219)

2026-08-08 00:29:36 +00:00

utils

[Spec Decode] DSpark confidence-scheduled verification (#47808)

2026-08-12 12:27:48 -07:00

v1

[Bugfix] Report FULL_ATTENTION for uniform-base UniformTypeKVCacheSpecs groups instead of UNKNOWN (#51218)

2026-08-13 01:25:23 +00:00

vllm_flash_attn

[Bugfix][Attention] Forward per-head FP8 descales through FA4 (#51363)

2026-08-11 14:05:20 -04:00

__init__.py

…

_aiter_ops.py

[ROCm] Enable DeepSeek-V4 on gfx11 (#47017)

2026-08-12 08:55:10 +08:00

_custom_ops.py

[Kernel][XPU] Tensor-descriptor operand loads for Triton W8A8 scaled_mm (#47205)

2026-08-10 00:56:11 +00:00

_xpu_ops.py

[Bugfix][XPU] Run GDN attention as eager break under breakable cudagraph (#51928)

2026-08-12 16:23:39 +08:00

collect_env.py

[XPU] fix collecting oneccl version info (#47104)

2026-08-04 11:53:28 +00:00

connections.py

…

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

[Spec Decode] DSpark confidence-scheduled verification (#47808)

2026-08-12 12:27:48 -07:00

exceptions.py

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665)

2026-07-29 04:20:40 +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

…

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

[Frontend][Core] Standardize request error handling with VLLMError hierarchy (#49665)

2026-07-29 04:20:40 +00:00

py.typed

…

sampling_params.py

[Spec Decode] DSpark confidence-scheduled verification (#47808)

2026-08-12 12:27:48 -07:00

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

…