Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

FEAT: Refactor device related code and add initial Intel GPU support #968

Merged
merged 5 commits into from
Feb 21, 2024
Merged

FEAT: Refactor device related code and add initial Intel GPU support #968

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
3b33d52
FEAT: Refactor device related code and add initial Intel GPU support
Feb 1, 2024
25378df
Fix lint
qinxuye Feb 5, 2024
74ad983
Fix lint
qinxuye Feb 5, 2024
16215bd
Move IPEX import earlier
Feb 6, 2024
2278ac0
Fix lint error due to rebase.
Feb 12, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions setup.cfg
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -95,6 +95,9 @@ all =
orjson
auto-gptq ; sys_platform!='darwin'
optimum
intel =
torch==2.1.0a0
intel_extension_for_pytorch==2.1.10+xpu
ggml =
llama-cpp-python>=0.2.25
ctransformers
Expand Down
6 changes: 6 additions & 0 deletions xinference/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -18,6 +18,12 @@
__version__ = _version.get_versions()["version"]


try:
import intel_extension_for_pytorch # noqa: F401
except:
pass


def _install():
from xoscar.backends.router import Router

Expand Down
176 changes: 110 additions & 66 deletions xinference/api/restful_api.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -231,9 +231,11 @@ def serve(self, logging_conf: Optional[dict] = None):
"/v1/ui/{model_uid}",
self.build_gradio_interface,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/token", self.login_for_access_token, methods=["POST"]
Expand All @@ -246,176 +248,218 @@ def serve(self, logging_conf: Optional[dict] = None):
"/v1/models/instances",
self.get_instance_info,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models/{model_type}/{model_name}/versions",
self.get_model_versions,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models",
self.list_models,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)

self._router.add_api_route(
"/v1/models/{model_uid}",
self.describe_model,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models/{model_uid}/events",
self.get_model_events,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models/instance",
self.launch_model_by_version,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models",
self.launch_model,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/experimental/speculative_llms",
self.launch_speculative_llm,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:start"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/models/{model_uid}",
self.terminate_model,
methods=["DELETE"],
dependencies=[Security(self._auth_service, scopes=["models:stop"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:stop"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/completions",
self.create_completion,
methods=["POST"],
response_model=Completion,
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/embeddings",
self.create_embedding,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/rerank",
self.rerank,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/audio/transcriptions",
self.create_transcriptions,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/audio/translations",
self.create_translations,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/images/generations",
self.create_images,
methods=["POST"],
response_model=ImageList,
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/images/variations",
self.create_variations,
methods=["POST"],
response_model=ImageList,
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/chat/completions",
self.create_chat_completion,
methods=["POST"],
response_model=ChatCompletion,
dependencies=[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:read"])]
if self.is_authenticated()
else None
),
)

# for custom models
self._router.add_api_route(
"/v1/model_registrations/{model_type}",
self.register_model,
methods=["POST"],
dependencies=[Security(self._auth_service, scopes=["models:register"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:register"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/model_registrations/{model_type}/{model_name}",
self.unregister_model,
methods=["DELETE"],
dependencies=[Security(self._auth_service, scopes=["models:unregister"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:unregister"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/model_registrations/{model_type}",
self.list_model_registrations,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)
self._router.add_api_route(
"/v1/model_registrations/{model_type}/{model_name}",
self.get_model_registrations,
methods=["GET"],
dependencies=[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None,
dependencies=(
[Security(self._auth_service, scopes=["models:list"])]
if self.is_authenticated()
else None
),
)

# Clear the global Registry for the MetricsMiddleware, or
Expand Down
5 changes: 3 additions & 2 deletions xinference/core/model.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -44,6 +44,7 @@

logger = logging.getLogger(__name__)

from ..device_utils import empty_cache
from .utils import json_dumps, log_async

try:
Expand Down Expand Up @@ -130,7 +131,7 @@ async def __pre_destroy__(self):
try:
import gc

import torch
import torch # noqa: F401
except ImportError:
error_message = "Failed to import module 'torch'"
installation_guide = [
Expand All @@ -141,7 +142,7 @@ async def __pre_destroy__(self):

del self._model
gc.collect()
torch.cuda.empty_cache()
empty_cache()

def __init__(
self,
Expand Down
6 changes: 3 additions & 3 deletions xinference/core/supervisor.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -227,11 +227,11 @@ async def get_builtin_families() -> Dict[str, List[str]]:
}

async def get_devices_count(self) -> int:
from ..utils import cuda_count
from ..device_utils import gpu_count

if self.is_local_deployment():
return cuda_count()
# distributed deployment, choose a worker and return its cuda_count.
return gpu_count()
# distributed deployment, choose a worker and return its device_count.
# Assume that each worker has the same count of cards.
worker_ref = await self._choose_worker()
return await worker_ref.get_devices_count()
Expand Down
Loading
Loading