HAIAI
aee76334d9
[amd_dev] branch rebase ( #25753 )
...
Signed-off-by: Nick Hill <nhill@redhat.com>
Signed-off-by: Lucas Kabela <lucaskabela@meta.com>
Signed-off-by: Max de Bayser <mbayser@br.ibm.com>
Signed-off-by: Andrew Sansom <andrew@protopia.ai>
Signed-off-by: Boyuan Feng <boyuan@meta.com>
Signed-off-by: Boyuan Feng <fby.1994@gmail.com>
Signed-off-by: boyuanfeng <boyuan@meta.com>
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Signed-off-by: JartX <sagformas@epdcenter.es>
Signed-off-by: Chendi Xue <Chendi.Xue@intel.com>
Signed-off-by: chaunceyjiang <chaunceyjiang@gmail.com>
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
Signed-off-by: Chen Zhang <zhangch99@outlook.com>
Signed-off-by: Roger Wang <hey@rogerw.io>
Signed-off-by: mgoin <mgoin64@gmail.com>
Signed-off-by: wwl2755 <wangwenlong2755@gmail.com>
Signed-off-by: Manoel Marques <manoel.marques@ibm.com>
Signed-off-by: Manoel Marques <manoelmrqs@gmail.com>
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: pengdrumli <pengdrumli@tencent.com>
Signed-off-by: windsonsea <haifeng.yao@daocloud.io>
Signed-off-by: Woosuk Kwon <woosuk@thinkingmachines.ai>
Signed-off-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Signed-off-by: Huamin Li <3ericli@gmail.com>
Signed-off-by: simondanielsson <simon.danielsson99@hotmail.com>
Signed-off-by: Rahul Tuli <rtuli@redhat.com>
Signed-off-by: Yang <lymailforjob@gmail.com>
Signed-off-by: Debolina Roy <debroy@redhat.com>
Signed-off-by: David Chen <530634352@qq.com>
Signed-off-by: wangzi <3220100013@zju.edu.cn>
Signed-off-by: Eldar Kurtic <8884008+eldarkurtic@users.noreply.github.com>
Signed-off-by: NickLucche <nlucches@redhat.com>
Signed-off-by: Yizhou Liu <liu_yizhou@outlook.com>
Signed-off-by: Sara Kokkila Schumacher <saraks@ibm.com>
Signed-off-by: Csrayz <jover@cmbchina.com>
Signed-off-by: ivyilike <pww123@cmbchina.com>
Signed-off-by: Burkhard Ringlein <ngl@zurich.ibm.com>
Signed-off-by: Bowen Wang <abmfy@icloud.com>
Signed-off-by: qqma <qqma@amazon.com>
Signed-off-by: ElizaWszola <ewszola@redhat.com>
Signed-off-by: Lu Fang <fanglu@fb.com>
Signed-off-by: Zhuohan Li <zhuohan123@gmail.com>
Signed-off-by: Luka Govedič <lgovedic@redhat.com>
Signed-off-by: luka <lgovedic@redhat.com>
Signed-off-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Signed-off-by: Or Ozeri <oro@il.ibm.com>
Signed-off-by: Johnny Yang <johnnyyang@google.com>
Signed-off-by: Alec Solder <alecs@fb.com>
Signed-off-by: Alec S <10566873+alecsolder@users.noreply.github.com>
Signed-off-by: Russell Bryant <rbryant@redhat.com>
Signed-off-by: Matthew Bonanni <mbonanni@redhat.com>
Signed-off-by: Alexander Matveev <amatveev@redhat.com>
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Signed-off-by: liuye.hj <liuye.hj@alibaba-inc.com>
Signed-off-by: Kunshang Ji <kunshang.ji@intel.com>
Signed-off-by: Lucia Fang <116399278+luccafong@users.noreply.github.com>
Signed-off-by: Michael Goin <mgoin64@gmail.com>
Signed-off-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Signed-off-by: Ming Yang <minos.future@gmail.com>
Signed-off-by: Zhikaiiii <1658973216@qq.com>
Signed-off-by: Andreas Hartel <andreas.hartel@aleph-alpha.com>
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
Signed-off-by: vllmellm <vllm.ellm@embeddedllm.com>
Signed-off-by: wuxibin <wuxibin@bytedance.com>
Signed-off-by: youkaichao <youkaichao@gmail.com>
Signed-off-by: Peter Pan <Peter.Pan@daocloud.io>
Signed-off-by: Peter Pan <peter.pan@daocloud.io>
Signed-off-by: Nicolò Lucchesi<nicolo.lucchesi@gmail.com>
Signed-off-by: Thomas Parnell <tpa@zurich.ibm.com>
Signed-off-by: Sage Moore <sage@neuralmagic.com>
Signed-off-by: Lucas Wilkinson <lwilkins@redhat.com>
Signed-off-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Signed-off-by: Tyler Michael Smith <tyler@neuralmagic.com>
Signed-off-by: Bill Nell <bnell@redhat.com>
Signed-off-by: Shreeasish Kumar <shreeasish@rivosinc.com>
Signed-off-by: Weida Hong <wdhongtw@google.com>
Signed-off-by: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.com>
Signed-off-by: Hashem Hashemi <hashem.hashemi@amd.com>
Signed-off-by: Hashem Hashemi <159079214+amd-hhashemi@users.noreply.github.com>
Signed-off-by: Amir Samani <asamani@nvidia.com>
Signed-off-by: ElizaWszola <elizaw.9289@gmail.com>
Signed-off-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
Signed-off-by: ilmarkov <markovilya197@gmail.com>
Signed-off-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Signed-off-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
Signed-off-by: rouchenzi <ruochenwen@gmail.com>
Signed-off-by: rouchenzi <40842833+rouchenzi@users.noreply.github.com>
Signed-off-by: Andrew Xia <axia@meta.com>
Signed-off-by: Kourosh Hakhamaneshi <kourosh@anyscale.com>
Signed-off-by: Corey Lowman <clowman1993@gmail.com>
Signed-off-by: jpvillam <jpvillam@amd.com>
Signed-off-by: dougbtv <dosmith@redhat.com>
Signed-off-by: Chenxi Yang <cxyang@fb.com>
Signed-off-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Signed-off-by: ahao-anyscale <ahao@anyscale.com>
Signed-off-by: Yan Lu <luyan@nvidia.com>
Signed-off-by: baxingpiaochong <771405853@qq.com>
Signed-off-by: Kyle Sayers <kylesayrs@gmail.com>
Signed-off-by: Nikhil Gupta <nikhil.gupta2@arm.com>
Signed-off-by: Yong Hoon Shin <yhshin@meta.com>
Signed-off-by: Benjamin Chislett <benjamin.chislett@centml.ai>
Signed-off-by: Benjamin Chislett <bchislett@nvidia.com>
Signed-off-by: Ben Browning <bbrownin@redhat.com>
Signed-off-by: Chengji Yao <chengjiyao@google.com>
Signed-off-by: jiang1.li <jiang1.li@intel.com>
Signed-off-by: Jackmin801 <ongjackm@gmail.com>
Signed-off-by: Jonas M. Kübler <44084297+jmkuebler@users.noreply.github.com>
Signed-off-by: taohui <taohui3@gmail.com>
Signed-off-by: rongfu.leng <rongfu.leng@daocloud.io>
Signed-off-by: Shu Wang <shuw@nvidia.com>
Signed-off-by: Shu Wang. <shuw@nvidia.com>
Signed-off-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Signed-off-by: Duncan Moss <djm.moss@gmail.com>
Signed-off-by: Shiyan Deng <dsy842974287@meta.com>
Signed-off-by: Wei Wei <wwei6@meta.com>
Signed-off-by: Saman Keon <samanamp@outlook.com>
Signed-off-by: yangxurui <yangxurui@meituan.com>
Signed-off-by: nicole-lihui <nicole.li@daocloud.io>
Signed-off-by: courage17340 <courage17340@163.com>
Signed-off-by: Jacob Kahn <jacobkahn1@gmail.com>
Signed-off-by: Fadi Arafeh <fadi.arafeh@arm.com>
Signed-off-by: Agata Dobrzyniewicz <adobrzyniewicz@habana.ai>
Signed-off-by: zxw <1020938856@qq.com>
Signed-off-by: wang.yuqi <noooop@126.com>
Signed-off-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Signed-off-by: chenlang <chen.lang5@zte.com.cn>
Signed-off-by: Jonas Kuebler <kuebj@amazon.com>
Signed-off-by: AlonKejzman <alonkeizman@gmail.com>
Signed-off-by: Tao Hui <taohui3@gmail.com>
Signed-off-by: Matthew Bonanni <mbonanni001@gmail.com>
Signed-off-by: Tomer Asida <57313761+tomeras91@users.noreply.github.com>
Signed-off-by: Aleksandr Malyshev <maleksan@amd.com>
Signed-off-by: Eugene Khvedchenia <ekhvedchenia@nvidia.com>
Signed-off-by: Eugene Khvedchenya <ekhvedchenya@gmail.com>
Signed-off-by: yiting.jiang <yiting.jiang@daocloud.io>
Signed-off-by: xaguilar <Xavier.AguilarFruto@amd.com>
Signed-off-by: Iceber Gu <caiwei95@hotmail.com>
Signed-off-by: Tao He <linzhu.ht@alibaba-inc.com>
Signed-off-by: Icey <1790571317@qq.com>
Signed-off-by: 许文卿 <xwq391974@alibaba-inc.com>
Signed-off-by: Chih-Chieh-Yang <7364402+cyang49@users.noreply.github.com>
Co-authored-by: Nick Hill <nhill@redhat.com>
Co-authored-by: Lucas Kabela <lucasakabela@gmail.com>
Co-authored-by: Maximilien de Bayser <mbayser@br.ibm.com>
Co-authored-by: Andrew Sansom <andrew@protopia.ai>
Co-authored-by: Boyuan Feng <boyuan@meta.com>
Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
Co-authored-by: JartX <sagformas@epdcenter.es>
Co-authored-by: Chendi.Xue <chendi.xue@intel.com>
Co-authored-by: Chauncey <chaunceyjiang@gmail.com>
Co-authored-by: xin.li <xin.li@daocloud.io>
Co-authored-by: Cyrus Leung <tlleungac@connect.ust.hk>
Co-authored-by: Chen Zhang <zhangch99@outlook.com>
Co-authored-by: Roger Wang <hey@rogerw.io>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: Wenlong Wang <wangwenlong2755@gmail.com>
Co-authored-by: Manoel Marques <manoelmrqs@gmail.com>
Co-authored-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Co-authored-by: lirong <56789630+lirong-lirong@users.noreply.github.com>
Co-authored-by: Michael Yao <haifeng.yao@daocloud.io>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Huamin Li <3ericli@gmail.com>
Co-authored-by: Lu Fang <30275821+houseroad@users.noreply.github.com>
Co-authored-by: Simon Danielsson <70206058+simondanielsson@users.noreply.github.com>
Co-authored-by: Rahul Tuli <rtuli@redhat.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Yang Liu <127183760+KKSK-DON@users.noreply.github.com>
Co-authored-by: Deboleina <debroy@redhat.com>
Co-authored-by: yinz-aizip <yinz@aizip.ai>
Co-authored-by: WeiQing Chen <40507679+david6666666@users.noreply.github.com>
Co-authored-by: wangzi <3220100013@zju.edu.cn>
Co-authored-by: Eldar Kurtić <8884008+eldarkurtic@users.noreply.github.com>
Co-authored-by: Nicolò Lucchesi <nlucches@redhat.com>
Co-authored-by: Ye (Charlotte) Qi <yeq@meta.com>
Co-authored-by: Yizhou <136800916+yiz-liu@users.noreply.github.com>
Co-authored-by: Sara-KS <50249410+Sara-KS@users.noreply.github.com>
Co-authored-by: Csrayz <jover@cmbchina.com>
Co-authored-by: ivyilike <pww123@cmbchina.com>
Co-authored-by: Burkhard Ringlein <ngl@zurich.ibm.com>
Co-authored-by: Bowen Wang <abmfy@icloud.com>
Co-authored-by: Tyler Michael Smith <tyler@neuralmagic.com>
Co-authored-by: Daisy-Ma-coder <daisy.ma.0117@gmail.com>
Co-authored-by: qqma <qqma@amazon.com>
Co-authored-by: ElizaWszola <ewszola@redhat.com>
Co-authored-by: Lucia Fang <116399278+luccafong@users.noreply.github.com>
Co-authored-by: Zhuohan Li <zhuohan123@gmail.com>
Co-authored-by: Simon Mo <simon.mo@hey.com>
Co-authored-by: Or Ozeri <oro@il.ibm.com>
Co-authored-by: Johnny Yang <24908445+jcyang43@users.noreply.github.com>
Co-authored-by: Chengji Yao <chengjiyao@google.com>
Co-authored-by: Alec S <10566873+alecsolder@users.noreply.github.com>
Co-authored-by: Alec Solder <alecs@fb.com>
Co-authored-by: Russell Bryant <rbryant@redhat.com>
Co-authored-by: Matthew Bonanni <mbonanni@redhat.com>
Co-authored-by: Robert Shaw <114415538+robertgshaw2-redhat@users.noreply.github.com>
Co-authored-by: Chris Bamford <chrisbam4d@gmail.com>
Co-authored-by: Alexander Matveev <59768536+alexm-redhat@users.noreply.github.com>
Co-authored-by: Wentao Ye <44945378+yewentao256@users.noreply.github.com>
Co-authored-by: JJJYmmm <92386084+JJJYmmm@users.noreply.github.com>
Co-authored-by: liuye.hj <liuye.hj@alibaba-inc.com>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: Lucia (Lu) Fang <fanglu@meta.com>
Co-authored-by: Varun Sundar Rabindranath <varunsundar08@gmail.com>
Co-authored-by: Varun Sundar Rabindranath <vsundarr@redhat.com>
Co-authored-by: Ming Yang <yming@meta.com>
Co-authored-by: Zhikaiiii <55917203+Zhikaiiii@users.noreply.github.com>
Co-authored-by: Andreas Hartel <andreas@hartel.me>
Co-authored-by: Jee Jee Li <pandaleefree@gmail.com>
Co-authored-by: vllmellm <vllm.ellm@embeddedllm.com>
Co-authored-by: Joel <wuxibin89@163.com>
Co-authored-by: youkaichao <youkaichao@gmail.com>
Co-authored-by: Mark McLoughlin <markmc@redhat.com>
Co-authored-by: Peter Pan <peter.pan@daocloud.io>
Co-authored-by: Nicolò Lucchesi <nicolo.lucchesi@gmail.com>
Co-authored-by: Fanli Lin <fanli.lin@intel.com>
Co-authored-by: Thomas Parnell <tpa@zurich.ibm.com>
Co-authored-by: Lucas Wilkinson <LucasWilkinson@users.noreply.github.com>
Co-authored-by: Sage Moore <sage@neuralmagic.com>
Co-authored-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: bnellnm <49004751+bnellnm@users.noreply.github.com>
Co-authored-by: rivos-shreeasish <shreeasish@rivosinc.com>
Co-authored-by: Chih-Chieh Yang <chih.chieh.yang@ibm.com>
Co-authored-by: Weida Hong <wdhongtw@gmail.com>
Co-authored-by: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.com>
Co-authored-by: Hashem Hashemi <159079214+amd-hhashemi@users.noreply.github.com>
Co-authored-by: Amir Samani <samani@ualberta.ca>
Co-authored-by: Luka Govedič <lgovedic@redhat.com>
Co-authored-by: jiahanc <173873397+jiahanc@users.noreply.github.com>
Co-authored-by: Ilya Markov <markovilya197@gmail.com>
Co-authored-by: Gregory Shtrasberg <156009573+gshtras@users.noreply.github.com>
Co-authored-by: Jialin Ouyang <Jialin.Ouyang@gmail.com>
Co-authored-by: rouchenzi <40842833+rouchenzi@users.noreply.github.com>
Co-authored-by: Andrew Xia <axia@meta.com>
Co-authored-by: kourosh hakhamaneshi <31483498+kouroshHakha@users.noreply.github.com>
Co-authored-by: Corey Lowman <clowman1993@gmail.com>
Co-authored-by: Juan Villamizar <100237675+jpvillam-amd@users.noreply.github.com>
Co-authored-by: jpvillam <jpvillam@amd.com>
Co-authored-by: Doug Smith <dosmith@redhat.com>
Co-authored-by: Chenxi Yang <cxyang@cs.utexas.edu>
Co-authored-by: Chenxi Yang <cxyang@fb.com>
Co-authored-by: ahao-anyscale <ahao@anyscale.com>
Co-authored-by: 0xNullPath <luyanfcp@foxmail.com>
Co-authored-by: baxingpiaochong <771405853@qq.com>
Co-authored-by: Benjamin Chislett <bchislett@nvidia.com>
Co-authored-by: Kyle Sayers <kylesayrs@gmail.com>
Co-authored-by: Nikhil Gupta <nikhil.gupta2@arm.com>
Co-authored-by: Yong Hoon Shin <48474650+sarckk@users.noreply.github.com>
Co-authored-by: lhsjohn <huashuoli@tencent.com>
Co-authored-by: Ben Browning <bbrownin@redhat.com>
Co-authored-by: Li, Jiang <jiang1.li@intel.com>
Co-authored-by: Jackmin801 <56836461+Jackmin801@users.noreply.github.com>
Co-authored-by: Jonas M. Kübler <44084297+jmkuebler@users.noreply.github.com>
Co-authored-by: Tao Hui <taohui3@gmail.com>
Co-authored-by: rongfu.leng <rongfu.leng@daocloud.io>
Co-authored-by: Shu Wang <shuw@nvidia.com>
Co-authored-by: Tyler Michael Smith <tlrmchlsmth@gmail.com>
Co-authored-by: Duncan Moss <djm.moss@gmail.com>
Co-authored-by: Shiyan Deng <dsy842974287@meta.com>
Co-authored-by: Wei Wei <wwei6@meta.com>
Co-authored-by: Saman A. Pour <samanamp@outlook.com>
Co-authored-by: XuruiYang <530534756@qq.com>
Co-authored-by: yangxurui <yangxurui@meituan.com>
Co-authored-by: Nicole LiHui 🥜 <nicolelihui@outlook.com>
Co-authored-by: courage17340 <courage17340@users.noreply.github.com>
Co-authored-by: Jacob Kahn <jacobkahn1@gmail.com>
Co-authored-by: Nicole LiHui 🥜 <nicole.li@daocloud.io>
Co-authored-by: Fadi Arafeh <115173828+fadara01@users.noreply.github.com>
Co-authored-by: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com>
Co-authored-by: yyzxw <34639446+yyzxw@users.noreply.github.com>
Co-authored-by: wang.yuqi <noooop@126.com>
Co-authored-by: Cyrus Leung <cyrus.tl.leung@gmail.com>
Co-authored-by: chenlang <chen.lang5@zte.com.cn>
Co-authored-by: chenlang <10346245@zte.com.cn>
Co-authored-by: AlonKejzman <alonkeizman@gmail.com>
Co-authored-by: tomeras91 <57313761+tomeras91@users.noreply.github.com>
Co-authored-by: Aleksandr Malyshev <164964928+maleksan85@users.noreply.github.com>
Co-authored-by: Aleksandr Malyshev <maleksan@amd.com>
Co-authored-by: Doug Lehr <douglehr@amd.com>
Co-authored-by: Eugene Khvedchenya <ekhvedchenya@gmail.com>
Co-authored-by: yitingdc <59356937+yitingdc@users.noreply.github.com>
Co-authored-by: xaguilar-amd <xavier.aguilarfruto@amd.com>
Co-authored-by: Iceber Gu <caiwei95@hotmail.com>
Co-authored-by: Tao He <linzhu.ht@alibaba-inc.com>
Co-authored-by: Icey <1790571317@qq.com>
Co-authored-by: Xu Wenqing <121550081+Xu-Wenqing@users.noreply.github.com>
Co-authored-by: Chih-Chieh Yang <7364402+cyang49@users.noreply.github.com>
Co-authored-by: RishiAstra <40644327+RishiAstra@users.noreply.github.com>
2025-09-26 17:14:31 +01:00
Wentao Ye
4771df7b2b
[Feature] Non-contiguous Support for FP8 Quantization ( #21961 )
...
Signed-off-by: yewentao256 <zhyanwentao@126.com>
Co-authored-by: mgoin <mgoin64@gmail.com>
2025-08-05 02:36:43 -07:00
Wentao Ye
1b0a155534
[Perf] Using __nv_fp8_e4m3 instead of c10::e4m3 for per_token_group_quant ( #21867 )
...
Signed-off-by: yewentao256 <zhyanwentao@126.com>
2025-07-29 21:50:46 -06:00
Wentao Ye
75d29cf4e1
[Perf] Cuda Kernel for Int8 Per Token Group Quant ( #21476 )
...
Signed-off-by: yewentao256 <zhyanwentao@126.com>
2025-07-25 17:07:07 -07:00
Wentao Ye
774d0c014b
[Perf] Cuda Kernel for Per Token Group Quant ( #21083 )
...
Signed-off-by: yewentao256 <zhyanwentao@126.com>
2025-07-22 07:27:15 -07:00
Mickaël Seznec
4fb56914c5
[perf] Add fused MLA QKV + strided layernorm ( #21116 )
...
Signed-off-by: Mickael Seznec <mickael@mistral.ai>
Co-authored-by: mgoin <mgoin64@gmail.com>
2025-07-22 07:07:44 -07:00
Lu Fang
c6703d1e0d
[MISC] Remove unused variableds in C++ ( #19609 )
...
Signed-off-by: Lu Fang <lufang@fb.com>
2025-06-15 20:05:28 -07:00
Michael Goin
e31446b6c8
[Perf] Tune scaled_fp8_quant by increasing vectorization ( #18844 )
...
Signed-off-by: mgoin <mgoin64@gmail.com>
2025-06-03 13:48:25 -07:00
Michael Goin
a17cef70ea
Removed unused marlin cuda code ( #17684 )
...
Signed-off-by: mgoin <mgoin64@gmail.com>
2025-05-06 17:59:47 -07:00
Charlie Fu
e85829450d
[Feature][ROCm]Enable fusion pass for torch.compile on ROCm ( #15050 )
...
Signed-off-by: charlifu <charlifu@amd.com>
2025-03-31 04:42:18 -07:00
Lu Fang
8c0d15d5c5
[Misc][Easy] Annotate unused vars in the csrc files ( #14798 )
...
Signed-off-by: Lu Fang <lufang@fb.com>
2025-03-15 12:40:09 +08:00
Jeff Daily
2a602b055a
forward fix PR 14245, restore build on ROCm 6.2 ( #14709 )
...
Signed-off-by: Jeff Daily <jeff.daily@amd.com>
2025-03-13 20:40:15 -07:00
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
ℍ𝕠𝕝𝕝𝕠𝕨 𝕄𝕒𝕟
a31614e386
[ROCm][Quantization][Kernel] Use FP8 FNUZ when OCP flag is 0 or undefined ( #13851 )
...
Signed-off-by: Hollow Man <hollowman@opensuse.org>
2025-02-27 10:39:10 +08:00
Gregory Shtrasberg
aabeb2688f
[ROCm][Quantization][Kernel] Using HIP FP8 header ( #12593 )
2025-02-25 00:39:59 -08:00
Gregory Shtrasberg
0023cd2b9d
[ROCm] MI300A compile targets deprecation ( #13560 )
2025-02-19 23:05:00 -08:00
Luka Govedič
30870b4f66
[torch.compile] Dynamic fp8 + rms_norm fusion ( #10906 )
...
Signed-off-by: luka <luka@neuralmagic.com>
Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
2024-12-13 03:19:23 +00:00
Luka Govedič
4f93dfe952
[torch.compile] Fuse RMSNorm with quant ( #9138 )
...
Signed-off-by: luka <luka@neuralmagic.com>
Co-authored-by: youkaichao <youkaichao@126.com>
2024-11-08 21:20:08 +00:00
Tyler Michael Smith
c3fab5f769
[Bugfix][Kernel] Prevent integer overflow in fp8 dynamic per-token quantize kernel ( #9425 )
2024-10-16 23:46:06 +00:00
Lucas Wilkinson
aeb37c2a72
[CI/Build] Per file CUDA Archs (improve wheel size and dev build times) ( #8845 )
2024-10-03 22:55:25 -04:00
Luka Govedič
7937009a7e
[Kernel] Replaced blockReduce[...] functions with cub::BlockReduce ( #7233 )
...
Co-authored-by: Michael Goin <michael@neuralmagic.com>
2024-08-21 20:18:00 -04:00
Charlie Fu
e837b624f2
[Feature][Hardware][Amd] Add fp8 Linear Layer for Rocm ( #7210 )
2024-08-16 10:06:30 -07:00
Tyler Michael Smith
6e4852ce28
[CI/Build] Suppress divide-by-zero and missing return statement warnings ( #7001 )
2024-08-05 16:00:01 -04:00
Tyler Michael Smith
cbbc904470
[Kernel] Squash a few more warnings ( #6914 )
2024-07-30 13:50:42 -04:00
Tyler Michael Smith
50704f52c4
[Bugfix][Kernel] Promote another index to int64_t ( #6838 )
2024-07-26 18:41:04 +00:00
Tyler Michael Smith
fea59c7712
[Bugfix][Kernel] Use int64_t for indices in fp8 quant kernels ( #6649 )
2024-07-22 14:08:30 -06:00
Alexander Matveev
396d92d5e0
[Kernel][Core] Add AWQ support to the Marlin kernel ( #6612 )
2024-07-21 19:41:42 -04:00
Varun Sundar Rabindranath
2e26564259
[ Kernel ] FP8 Dynamic Per Token Quant - Add scale_ub ( #6593 )
...
Co-authored-by: Varun Sundar Rabindranth <varun@neuralmagic.com>
2024-07-19 18:15:26 -07:00
Varun Sundar Rabindranath
b5241e41d9
[ Kernel ] FP8 Dynamic-Per-Token Quant Kernel ( #6511 )
...
Co-authored-by: Varun Sundar Rabindranath <varun@neuralmagic.com>
2024-07-18 01:38:35 +00:00
Michael Goin
47f0954af0
[Kernel] Expand FP8 support to Ampere GPUs using FP8 Marlin ( #5975 )
2024-07-03 17:38:00 +00:00
Cody Yu
5985e3427d
[Kernel] Vectorized FP8 quantize kernel ( #5396 )
...
Inspired by #5146 , this PR improves FP8 quantize kernel by vectorizing data transfer to better utilize memory bandwidth. Microbenchmark shows that this improved kernel can achieve 1.0x-1.5x speedup (especially when hidden size is large).
In details, we applied 3 optimizations:
- Use inverted scale so that most divisions are changed to multiplications.
- Unroll the loop by 4 times to improve ILP.
- Use vectorized 4 to transfer data between HBM and SRAM.
2024-06-12 14:07:26 -07:00
bnellnm
5467ac3196
[Kernel][Misc] Use TORCH_LIBRARY instead of PYBIND11_MODULE for custom ops ( #5047 )
2024-06-09 16:23:30 -04:00
Michael Goin
5f6d10c14c
[CI/Build] Enforce style for C++ and CUDA code with clang-format ( #4722 )
2024-05-22 07:18:41 +00:00
Cody Yu
c833101740
[Kernel] Refactor FP8 kv-cache with NVIDIA float8_e4m3 support ( #4535 )
2024-05-09 18:04:17 -06:00
Philipp Moritz
a98187cf72
[Kernel] Make static FP8 scaling more robust ( #4570 )
...
Previously FP8 static scaling works if the scales are overestimating the maxima of all activation tensors during computation. However this will not always be the case even if the scales were calibrated very carefully. For example, with the activations in my checkpoint
https://huggingface.co/pcmoritz/Mixtral-8x7B-v0.1-fp8-act-scale
(which was calibrated on https://huggingface.co/datasets/HuggingFaceH4/ultrachat_200k ), I'm getting the following mostly random performance on MMLU:
| Groups |Version|Filter|n-shot|Metric|Value | |Stderr|
|------------------|-------|------|-----:|------|-----:|---|-----:|
|mmlu |N/A |none | 0|acc |0.2295|± |0.0035|
| - humanities |N/A |none | 5|acc |0.2421|± |0.0062|
| - other |N/A |none | 5|acc |0.2398|± |0.0076|
| - social_sciences|N/A |none | 5|acc |0.2171|± |0.0074|
| - stem |N/A |none | 5|acc |0.2125|± |0.0073|
With the fix in this PR where the scaled activations are clamped between [-std::numeric_limits<c10::Float8_e4m3fn>::max(), std::numeric_limits<c10::Float8_e4m3fn>::max()] to make sure there are no NaNs, the performance is
| Groups |Version|Filter|n-shot|Metric|Value | |Stderr|
|------------------|-------|------|-----:|------|-----:|---|-----:|
|mmlu |N/A |none | 0|acc |0.7008|± |0.0036|
| - humanities |N/A |none | 5|acc |0.6453|± |0.0065|
| - other |N/A |none | 5|acc |0.7692|± |0.0072|
| - social_sciences|N/A |none | 5|acc |0.8083|± |0.0070|
| - stem |N/A |none | 5|acc |0.6115|± |0.0083|
This is not perfect yet but is getting very close to the FP16 / dynamic activation scale performance.
2024-05-06 17:39:28 -07:00
Philipp Moritz
12628d3c78
[Kernel] Optimize FP8 support for MoE kernel / Mixtral via static scales ( #4343 )
...
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
2024-04-27 04:49:59 +00:00
Philipp Moritz
eace8bf0b9
[Kernel] FP8 support for MoE kernel / Mixtral ( #4244 )
...
This PR is the first step towards fixing https://github.com/vllm-project/vllm/pull/3208
It implements dynamic per-tensor scaling (see https://github.com/vllm-project/vllm/pull/4118 ), so users do not need to compute activation scales on a calibration dataset and they also don't need to convert their model checkpoints. It is enough to specify the `quantization="fp8"` argument. You can try out the PR like this:
```python
from vllm import LLM, SamplingParams
prompts = [
"Hello, my name is",
"The president of the United States is",
"The capital of France is",
"The future of AI is",
]
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
llm = LLM(model="mistralai/Mixtral-8x7B-Instruct-v0.1", tensor_parallel_size=2, quantization="fp8")
outputs = llm.generate(prompts, sampling_params)
# Print the outputs.
for output in outputs:
prompt = output.prompt
generated_text = output.outputs[0].text
print(f"Prompt: {prompt!r}, Generated text: {generated_text!r}")
```
**Performance**: For this PR, the focus is on making the code clean (while still trying to get reasonable performance), there is a bunch of optimizations that we will submit as a follow up PR that significantly improve the performance (similar to the numbers in https://github.com/vllm-project/vllm/pull/3954 ). With this PR, the results are as follows:
<img width="725" alt="Screenshot 2024-04-21 at 1 31 50 PM" src="https://github.com/vllm-project/vllm/assets/113316/d8fe1118-07a0-4d4e-8530-37a77d465a03 ">
**Accuracy**: The accuracy with this PR on MMLU on `mistralai/Mixtral-8x7B-v0.1` is as follows:
```
| Groups |Version|Filter|n-shot|Metric|Value | |Stderr|
|------------------|-------|------|-----:|------|-----:|---|-----:|
|mmlu |N/A |none | 0|acc |0.7018|± |0.0036|
| - humanities |N/A |none | 5|acc |0.6472|± |0.0065|
| - other |N/A |none | 5|acc |0.7673|± |0.0072|
| - social_sciences|N/A |none | 5|acc |0.8099|± |0.0070|
| - stem |N/A |none | 5|acc |0.6131|± |0.0083|
```
this compares favorably with the fp16 results which are
```
| Groups |Version|Filter|n-shot|Metric|Value | |Stderr|
|------------------|-------|------|-----:|------|-----:|---|-----:|
|mmlu |N/A |none | 0|acc |0.7020|± |0.1313|
| - humanities |N/A |none | 5|acc |0.6425|± |0.1349|
| - other |N/A |none | 5|acc |0.7744|± |0.1038|
| - social_sciences|N/A |none | 5|acc |0.8131|± |0.0695|
| - stem |N/A |none | 5|acc |0.6108|± |0.1383|
```
Happy hacking!
2024-04-24 01:18:23 +00:00
Adrian Abeyta
2ff767b513
Enable scaled FP8 (e4m3fn) KV cache on ROCm (AMD GPU) ( #3290 )
...
Co-authored-by: Gregory Shtrasberg <Gregory.Shtrasberg@amd.com>
Co-authored-by: HaiShaw <hixiao@gmail.com>
Co-authored-by: AdrianAbeyta <Adrian.Abeyta@amd.com>
Co-authored-by: Matthew Wong <Matthew.Wong2@amd.com>
Co-authored-by: root <root@gt-pla-u18-08.pla.dcgpu>
Co-authored-by: mawong-amd <156021403+mawong-amd@users.noreply.github.com>
Co-authored-by: ttbachyinsda <ttbachyinsda@outlook.com>
Co-authored-by: guofangze <guofangze@kuaishou.com>
Co-authored-by: Michael Goin <mgoin64@gmail.com>
Co-authored-by: jacobthebanana <50071502+jacobthebanana@users.noreply.github.com>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
2024-04-03 14:15:55 -07:00
