[Hardware][Intel-Gaudi] Add Intel Gaudi (HPU) inference backend (#6143)

Signed-off-by: yuwenzho <yuwen.zhou@intel.com>
Signed-off-by: Chendi.Xue <chendi.xue@intel.com>
Signed-off-by: Bob Zhu <bob.zhu@intel.com>
Signed-off-by: zehao-intel <zehao.huang@intel.com>
Signed-off-by: Konrad Zawora <kzawora@habana.ai>
Co-authored-by: Kunshang Ji <kunshang.ji@intel.com>
Co-authored-by: Sanju C Sudhakaran <scsudhakaran@habana.ai>
Co-authored-by: Michal Adamczyk <madamczyk@habana.ai>
Co-authored-by: Marceli Fylcek <mfylcek@habana.ai>
Co-authored-by: Himangshu Lahkar <49579433+hlahkar@users.noreply.github.com>
Co-authored-by: Vivek Goel <vgoel@habana.ai>
Co-authored-by: yuwenzho <yuwen.zhou@intel.com>
Co-authored-by: Dominika Olszewska <dolszewska@habana.ai>
Co-authored-by: barak goldberg <149692267+bgoldberg-habana@users.noreply.github.com>
Co-authored-by: Michal Szutenberg <37601244+szutenberg@users.noreply.github.com>
Co-authored-by: Jan Kaniecki <jkaniecki@habana.ai>
Co-authored-by: Agata Dobrzyniewicz <160237065+adobrzyniewicz-habana@users.noreply.github.com>
Co-authored-by: Krzysztof Wisniewski <kwisniewski@habana.ai>
Co-authored-by: Dudi Lester <160421192+dudilester@users.noreply.github.com>
Co-authored-by: Ilia Taraban <tarabanil@gmail.com>
Co-authored-by: Chendi.Xue <chendi.xue@intel.com>
Co-authored-by: Michał Kuligowski <mkuligowski@habana.ai>
Co-authored-by: Jakub Maksymczuk <jmaksymczuk@habana.ai>
Co-authored-by: Tomasz Zielinski <85164140+tzielinski-habana@users.noreply.github.com>
Co-authored-by: Sun Choi <schoi@habana.ai>
Co-authored-by: Iryna Boiko <iboiko@habana.ai>
Co-authored-by: Bob Zhu <41610754+czhu15@users.noreply.github.com>
Co-authored-by: hlin99 <73271530+hlin99@users.noreply.github.com>
Co-authored-by: Zehao Huang <zehao.huang@intel.com>
Co-authored-by: Andrzej Kotłowski <Andrzej.Kotlowski@intel.com>
Co-authored-by: Yan Tomsinsky <73292515+Yantom1@users.noreply.github.com>
Co-authored-by: Nir David <ndavid@habana.ai>
Co-authored-by: Yu-Zhou <yu.zhou@intel.com>
Co-authored-by: Ruheena Suhani Shaik <rsshaik@habana.ai>
Co-authored-by: Karol Damaszke <kdamaszke@habana.ai>
Co-authored-by: Marcin Swiniarski <mswiniarski@habana.ai>
Co-authored-by: Woosuk Kwon <woosuk.kwon@berkeley.edu>
Co-authored-by: Jacek Czaja <jacek.czaja@intel.com>
Co-authored-by: Jacek Czaja <jczaja@habana.ai>
Co-authored-by: Yuan <yuan.zhou@outlook.com>
This commit is contained in:
Konrad Zawora
2024-11-06 01:09:10 -08:00
committed by GitHub
co-authored by Kunshang Ji Sanju C Sudhakaran Michal Adamczyk Marceli Fylcek Himangshu Lahkar Vivek Goel yuwenzho Dominika Olszewska barak goldberg Michal Szutenberg Jan Kaniecki Agata Dobrzyniewicz Krzysztof Wisniewski Dudi Lester Ilia Taraban Chendi.Xue Michał Kuligowski Jakub Maksymczuk Tomasz Zielinski Sun Choi Iryna Boiko Bob Zhu hlin99 Zehao Huang Andrzej Kotłowski Yan Tomsinsky Nir David Yu-Zhou Ruheena Suhani Shaik Karol Damaszke Marcin Swiniarski Woosuk Kwon Jacek Czaja Jacek Czaja Yuan
parent a5fda50a10
commit a02a50e6e5
31 changed files with 4279 additions and 20 deletions
+19
View File
@@ -177,6 +177,7 @@ class GroupCoordinator:
use_pynccl: bool,
use_custom_allreduce: bool,
use_tpu_communicator: bool,
use_hpu_communicator: bool,
use_message_queue_broadcaster: bool = False,
group_name: Optional[str] = None,
):
@@ -213,6 +214,7 @@ class GroupCoordinator:
self.use_pynccl = use_pynccl
self.use_custom_allreduce = use_custom_allreduce
self.use_tpu_communicator = use_tpu_communicator
self.use_hpu_communicator = use_hpu_communicator
# lazy import to avoid documentation build error
from vllm.distributed.device_communicators.custom_all_reduce import (
@@ -241,6 +243,12 @@ class GroupCoordinator:
if use_tpu_communicator and self.world_size > 1:
self.tpu_communicator = TpuCommunicator(group=self.cpu_group)
from vllm.distributed.device_communicators.hpu_communicator import (
HpuCommunicator)
self.hpu_communicator: Optional[HpuCommunicator]
if use_hpu_communicator and self.world_size > 1:
self.hpu_communicator = HpuCommunicator(group=self.device_group)
from vllm.distributed.device_communicators.shm_broadcast import (
MessageQueue)
self.mq_broadcaster: Optional[MessageQueue] = None
@@ -362,6 +370,10 @@ class GroupCoordinator:
# TPU handles Dynamo with its own logic.
return self.tpu_communicator.all_reduce(input_)
if self.hpu_communicator is not None and \
not self.hpu_communicator.disabled:
return self.hpu_communicator.all_reduce(input_)
if self.ca_comm is not None and \
not self.ca_comm.disabled and \
self.ca_comm.should_custom_ar(input_):
@@ -400,6 +412,11 @@ class GroupCoordinator:
if tpu_comm is not None and not tpu_comm.disabled:
return tpu_comm.all_gather(input_, dim)
# For HPUs, use HPU communicator.
hpu_comm = self.hpu_communicator
if hpu_comm is not None and not hpu_comm.disabled:
return hpu_comm.all_gather(input_, dim)
if dim < 0:
# Convert negative dim to positive.
dim += input_.dim()
@@ -879,6 +896,7 @@ def init_world_group(ranks: List[int], local_rank: int,
use_pynccl=False,
use_custom_allreduce=False,
use_tpu_communicator=False,
use_hpu_communicator=False,
group_name="world",
)
@@ -900,6 +918,7 @@ def init_model_parallel_group(
use_pynccl=True,
use_custom_allreduce=use_custom_allreduce,
use_tpu_communicator=True,
use_hpu_communicator=True,
use_message_queue_broadcaster=use_message_queue_broadcaster,
group_name=group_name,
)