Michael Goin
bcdfb2a330
[Bugfix] Fix incorrect dispatch for CutlassBlockScaledGroupedGemm and DeepGEMM ( #20933 )
...
Signed-off-by: mgoin <mgoin64@gmail.com >
2025-07-15 01:42:17 +00:00
Luka Govedič and mgoin
31d5c1797f
[Perf][fp8] Use CustomOp abstraction for fp8 quant for better perf ( #19830 )
...
Signed-off-by: Luka Govedic <lgovedic@redhat.com >
Co-authored-by: mgoin <mgoin64@gmail.com >
2025-07-11 04:56:28 +00:00
Wentao Ye
e2de455c34
[Feature] Integrate SM100 DeepGEMM support ( #20087 )
2025-07-10 20:18:05 -07:00
Duncan Moss and Duncan Moss
3d184b95b8
[feat]: CUTLASS block scaled group gemm for SM100 ( #19757 )
...
Signed-off-by: Duncan Moss <djm.moss@gmail.com >
Co-authored-by: Duncan Moss <dmoss@nvidia.com >
2025-07-04 12:58:04 -06:00
Michael Goin
0e3fe896e2
Support Llama 4 for fused_marlin_moe ( #20457 )
...
Signed-off-by: mgoin <mgoin64@gmail.com >
2025-07-04 07:55:10 +00:00
bnellnm
78fe77534b
[Kernel] Enable fp8 support for pplx and BatchedTritonExperts. ( #18864 )
...
Signed-off-by: Bill Nell <bnell@redhat.com >
2025-07-03 14:55:40 -07:00
bnellnm
2e25bb12a8
[Bugfix] Fix import of CutlassExpertsFp8 in compressed_tensors_moe.py ( #20381 )
...
Signed-off-by: Bill Nell <bnell@redhat.com >
2025-07-03 02:07:43 +00:00
bnellnm and ElizaWszola
c1909e7e8c
[Kernels] MoE refactor ( #19636 )
...
Signed-off-by: Bill Nell <bnell@redhat.com >
Signed-off-by: ElizaWszola <ewszola@redhat.com >
Co-authored-by: ElizaWszola <ewszola@redhat.com >
2025-07-02 06:08:27 -07:00
TJian
02cabff207
[V1] [ROCm] Enable EP with AITER Fused MoE ( #20270 )
...
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com >
2025-07-01 16:48:30 +00:00
Kyle Sayers
9025a9a705
[Quant] [Bugfix] Fix quantization config matching with hf_to_vllm_mapper ( #20046 )
2025-07-01 19:20:34 +09:00
Wentao Ye
4d36693687
[Refactor] Create a function util and cache the results for has_deepgemm, has_deepep, has_pplx ( #20187 )
...
Signed-off-by: yewentao256 <zhyanwentao@126.com >
2025-06-28 22:06:38 +00:00
Bowen Wang
e9fd658a73
[Feature] Expert Parallelism Load Balancer (EPLB) ( #18343 )
...
Signed-off-by: Bowen Wang <abmfy@icloud.com >
2025-06-26 15:30:21 -07:00
artetaout
b8e809a057
[Kernel] Support deep_gemm for linear methods ( #19085 )
...
Signed-off-by: artetaout <lulala341@gmail.com >
2025-06-11 15:14:45 +08:00
ElizaWszola and Tyler Michael Smith
84166fee97
[Kernel] Integrate CUTLASS MoE kernel with PPLX ( #18762 )
...
Signed-off-by: ElizaWszola <ewszola@redhat.com >
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com >
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com >
2025-06-06 18:26:11 -07:00
Varun Sundar Rabindranath and Varun
c3fd4d669a
[Kernel] Integrate batched/masked deepgemm kernel ( #19111 )
...
Signed-off-by: Varun <vsundarr@redhat.com >
Co-authored-by: Varun <vsundarr@redhat.com >
2025-06-04 21:59:18 +00:00
Varun Sundar Rabindranath and Varun Sundar Rabindranath
fa98d77773
[Kernel] DeepEP dispatch-combine kernel integration ( #18434 )
...
Signed-off-by: Varun <vsundarr@redhat.com >
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com >
2025-06-03 12:30:02 -07:00
Simon Mo
02f0c7b220
[Misc] Add SPDX-FileCopyrightText ( #19100 )
...
Signed-off-by: simon-mo <simon.mo@hey.com >
2025-06-03 11:20:17 -07:00
vllmellm
d260f799a9
[FEAT] [ROCm] Upgrade AITER Fused MoE kernels. ( #18271 )
...
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com >
2025-05-26 23:14:07 -07:00
Cyrus Leung
503f8487c2
[Misc] Reduce logs on startup ( #18649 )
...
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk >
2025-05-24 23:03:53 -07:00
youkaichao
6a7988c55b
Refactor pplx init logic to make it modular (prepare for deepep) ( #18200 )
...
Signed-off-by: youkaichao <youkaichao@gmail.com >
2025-05-23 23:43:43 +08:00
bnellnm
f9c069c85e
Modularize fused experts and integrate PPLX kernels ( #15956 )
2025-05-14 13:11:54 -07:00
TJian and Gregory Shtrasberg
612c2edb4f
[FEAT] [ROCm]: Add AITER CK 2 Stages MoE support ( #17110 )
...
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com >
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com >
2025-05-14 03:03:11 -07:00
Michael Goin
9a2a6357de
[Bugfix] Fix FP8 Marlin MoE and enable for compressed-tensors models ( #18026 )
...
Signed-off-by: mgoin <mgoin64@gmail.com >
2025-05-13 19:48:33 -07:00
vllmellm and tjtanaa
40de1ef455
[FEAT] [ROCm]: Add AITER Block-Scaled GEMM Feature ( #14968 )
...
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com >
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com >
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com >
2025-05-13 19:08:20 -07:00
Harry Mellor
6223dd8114
Update deprecated type hinting in model_executor/layers ( #18056 )
...
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com >
2025-05-13 04:17:23 -07:00
Jinzhen Lin
1d0c9d6b2d
[Kernel] some optimizations for dense marlin and moe marlin ( #16850 )
...
Signed-off-by: Jinzhen Lin <linjinzhen@hotmail.com >
2025-05-05 09:39:30 -07:00
Harry Mellor
13698db634
Improve configs - ModelConfig ( #17130 )
...
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com >
2025-04-30 10:38:22 +08:00
rasmith and Luka Govedič
a41351f363
[Quantization][FP8] Add support for FP8 models with input_scale for output projection and QK quantization ( #15734 )
...
Signed-off-by: Randall Smith <Randall.Smith@amd.com >
Signed-off-by: Luka Govedič <lgovedic@redhat.com >
Co-authored-by: Luka Govedič <lgovedic@redhat.com >
2025-04-25 00:45:02 -07:00
5b794cae8d
[ROCm] Add aiter tkw1 kernel for Llama4 fp8 ( #16727 )
...
Signed-off-by: kliuae <kuanfu.liu@embeddedllm.com >
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com >
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com >
Co-authored-by: vllmellm <vllm.ellm@embeddedllm.com >
2025-04-21 20:42:34 -07:00
Isotr0py
40b4284fe3
[Bugfix] Handle process_weights_after_loading for QKVCrossParallelLinear ( #15328 )
...
Signed-off-by: Isotr0py <2037008807@qq.com >
2025-04-08 10:02:23 -07:00
Michael Goin
21802c4b6d
[ROCm][Bugfix][FP8] Make fp8 quant respect fused modules mapping ( #16031 )
...
Signed-off-by: mgoin <michael@neuralmagic.com >
2025-04-07 21:28:14 -04:00
55dcce91df
Upstream Llama4 Support to Main ( #16113 )
...
Signed-off-by: Aston Zhang <22279212+astonzhang@users.noreply.github.com >
Signed-off-by: Chris Thi <chris.c.thi@gmail.com >
Signed-off-by: drisspg <drisspguessous@gmail.com >
Signed-off-by: Jon Swenson <jmswen@gmail.com >
Signed-off-by: Keyun Tong <tongkeyun@gmail.com >
Signed-off-by: Lu Fang <fanglu@meta.com >
Signed-off-by: Xiaodong Wang <xdwang@meta.com >
Signed-off-by: Yang Chen <yangche@fb.com >
Signed-off-by: Ye (Charlotte) Qi <yeq@meta.com >
Signed-off-by: Yong Hoon Shin <yhshin@meta.com >
Signed-off-by: Zijing Liu <liuzijing2014@gmail.com >
Signed-off-by: Lu Fang <lufang@fb.com >
Signed-off-by: Lu Fang <fanglu@fb.com >
Signed-off-by: Lucia Fang <fanglu@fb.com >
Signed-off-by: Roger Wang <ywang@roblox.com >
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk >
Co-authored-by: Lu Fang <fanglu@fb.com >
Co-authored-by: Roger Wang <ywang@roblox.com >
Co-authored-by: DarkLight1337 <tlleungac@connect.ust.hk >
2025-04-07 08:06:27 -07:00
bnellnm
e59ca942f5
Add option to use DeepGemm contiguous grouped gemm kernel for fused MoE operations. ( #13932 )
...
Signed-off-by: Bill Nell <bnell@redhat.com >
2025-04-01 12:07:43 -04:00
Gregory Shtrasberg and Luka Govedič
4d0ec37267
[Quantization][FP8] Adding support for fp8 gemm layer input in fp8 ( #14578 )
...
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com >
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com >
2025-03-28 02:58:16 +00:00
Robert Shaw
e1e0fd7543
[TPU] Avoid Triton Import ( #15589 )
...
Signed-off-by: rshaw@neuralmagic.com <robertgshaw2@gmail.com >
2025-03-27 06:43:02 +00:00
vllmellm and tjtanaa
5ebf66748b
[FEAT][ROCm] Integrate Fused MoE Kernels from AITER ( #14967 )
...
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com >
Signed-off-by: tjtanaa <tunjian.tan@embeddedllm.com >
Co-authored-by: tjtanaa <tunjian.tan@embeddedllm.com >
2025-03-26 16:30:30 +08:00
Gregory Shtrasberg and charlifu
f533b5837f
[ROCm][Kernel] MoE weights padding ( #14454 )
...
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com >
Signed-off-by: charlifu <charlifu@amd.com >
Co-authored-by: charlifu <charlifu@amd.com >
2025-03-24 23:45:30 +00:00
Jeff Daily
a1c8f3796c
dynamic distpatch of fp8 kernels ( #14245 )
...
Signed-off-by: Jeff Daily <jeff.daily@amd.com >
2025-03-11 10:54:56 -04:00
Luka Govedič
e1744502c2
[FP8] Refactor apply_fp8_linear and apply_fp8_linear_generic into an object ( #14390 )
...
Signed-off-by: luka <luka@neuralmagic.com >
2025-03-07 05:20:16 +00:00
Woosuk Kwon
b382a7f28f
[BugFix] Make FP8 Linear compatible with torch.compile ( #13918 )
...
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu >
2025-02-26 13:48:55 -08:00
Michael Goin
07c4353057
[Model] Support Grok1 ( #13795 )
...
Signed-off-by: mgoin <mgoin64@gmail.com >
2025-02-26 01:07:12 +00:00
Tyler Michael Smith
1e15aaef56
[Bugfix][Quantization] Fix FP8 + EP ( #13784 )
...
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com >
2025-02-25 10:54:17 +08:00
Jongseok Park
781096e385
Expert Parallelism (EP) Support for DeepSeek V2 ( #12583 )
2025-02-24 07:33:20 -08:00
Gregory Shtrasberg
c904fdddf6
[ROCm] Apply FP8 weights padding to values not divisible by 512 bytes on ROCm ( #13231 )
2025-02-22 05:54:38 -08:00
Tyler Michael Smith
b3942e157e
[Bugfix][CI][V1] Work around V1 + CUDA Graph + torch._scaled_mm fallback issue ( #13425 )
...
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com >
2025-02-18 00:32:48 +00:00
Kyle Sayers
12913d17ba
[Quant] Add SupportsQuant to phi3 and clip ( #13104 )
2025-02-15 19:28:33 -08:00
Russell Bryant
e489ad7a21
[Misc] Add SPDX-License-Identifier headers to python source files ( #12628 )
...
- **Add SPDX license headers to python source files**
- **Check for SPDX headers using pre-commit**
commit 9d7ef44c3cfb72ca4c32e1c677d99259d10d4745
Author: Russell Bryant <rbryant@redhat.com >
Date: Fri Jan 31 14:18:24 2025 -0500
Add SPDX license headers to python source files
This commit adds SPDX license headers to python source files as
recommended to
the project by the Linux Foundation. These headers provide a concise way
that is
both human and machine readable for communicating license information
for each
source file. It helps avoid any ambiguity about the license of the code
and can
also be easily used by tools to help manage license compliance.
The Linux Foundation runs license scans against the codebase to help
ensure
we are in compliance with the licenses of the code we use, including
dependencies. Having these headers in place helps that tool do its job.
More information can be found on the SPDX site:
- https://spdx.dev/learn/handling-license-info/
Signed-off-by: Russell Bryant <rbryant@redhat.com >
commit 5a1cf1cb3b80759131c73f6a9dddebccac039dea
Author: Russell Bryant <rbryant@redhat.com >
Date: Fri Jan 31 14:36:32 2025 -0500
Check for SPDX headers using pre-commit
Signed-off-by: Russell Bryant <rbryant@redhat.com >
---------
Signed-off-by: Russell Bryant <rbryant@redhat.com >
2025-02-02 11:58:18 -08:00
Tyler Michael Smith
eb5741ad42
[Kernel][Quantization] Integrate block-quantized CUTLASS kernels for DeepSeekV3 ( #12587 )
...
Integrates the block-quantized kernels introduced in
https://github.com/vllm-project/vllm/pull/11868 for use in linear
layers.
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com >
2025-01-31 15:29:11 -08:00
Robert Shaw and simon-mo
325f679f32
[BugFix] Fix Torch.Compile For DeepSeek ( #12594 )
...
Co-authored-by: simon-mo <xmo@berkeley.edu >
2025-01-31 12:06:39 -08:00
eb5cb5e528
[BugFix] Fix parameter names and process_after_weight_loading for W4A16 MoE Group Act Order ( #11528 )
...
Signed-off-by: ElizaWszola <eliza@neuralmagic.com >
Co-authored-by: ElizaWszola <eliza@neuralmagic.com >
Co-authored-by: Michael Goin <michael@neuralmagic.com >
2025-01-23 21:40:33 +00:00
Michael Goin