[LLM] Fix linux load libs for NeoX and llama (#8257)

* init

* add lisence

* fix style
This commit is contained in:
Yina Chen 2023-06-02 17:03:17 +08:00 committed by GitHub
parent 286b010bf1
commit 657ea0ee50
5 changed files with 47 additions and 11 deletions

View file

@ -63,28 +63,29 @@ from ctypes import (
)
import pathlib
from bigdl.llm.utils.common import invalidInputError
from bigdl.llm.utils import get_avx_flags
# Load the library
def _load_shared_library(lib_base_name: str):
# Determine the file extension based on the platform
if sys.platform.startswith("linux"):
lib_ext = ".so"
elif sys.platform == "darwin":
if sys.platform.startswith("linux") or sys.platform == "darwin":
lib_ext = ".so"
elif sys.platform == "win32":
lib_ext = ".dll"
else:
invalidInputError(False, "Unsupported platform.")
avx = get_avx_flags()
# Construct the paths to the possible shared library names (python/llm/src/bigdl/llm/libs)
_base_path = pathlib.Path(__file__).parent.parent.parent.parent.resolve()
_base_path = _base_path / 'libs'
# Searching for the library in the current directory under the name "libgptneox" (default name
# for gptneoxcpp) and "gptneox" (default name for this repo)
_lib_paths = [
_base_path / f"lib{lib_base_name}{lib_ext}",
_base_path / f"{lib_base_name}{lib_ext}",
_base_path / f"lib{lib_base_name}{avx}{lib_ext}",
_base_path / f"{lib_base_name}{avx}{lib_ext}",
]
if "GPTNEOX_CPP_LIB" in os.environ:

View file

@ -63,28 +63,29 @@ from ctypes import (
)
import pathlib
from bigdl.llm.utils.common import invalidInputError
from bigdl.llm.utils import get_avx_flags
# Load the library
def _load_shared_library(lib_base_name: str):
# Determine the file extension based on the platform
if sys.platform.startswith("linux"):
lib_ext = ".so"
elif sys.platform == "darwin":
if sys.platform.startswith("linux") or sys.platform == "darwin":
lib_ext = ".so"
elif sys.platform == "win32":
lib_ext = ".dll"
else:
invalidInputError(False, "Unsupported platform.")
avx = get_avx_flags()
# Construct the paths to the possible shared library names
_base_path = pathlib.Path(__file__).parent.parent.parent.parent.resolve()
_base_path = _base_path / 'libs'
# Searching for the library in the current directory under the name "libllama" (default name
# for llamacpp) and "llama" (default name for this repo)
_lib_paths = [
_base_path / f"lib{lib_base_name}{lib_ext}",
_base_path / f"{lib_base_name}{lib_ext}",
_base_path / f"lib{lib_base_name}{avx}{lib_ext}",
_base_path / f"{lib_base_name}{avx}{lib_ext}",
]
if "LLAMA_CPP_LIB" in os.environ:

View file

@ -18,3 +18,5 @@
# physically located elsewhere.
# Otherwise there would be module not found error in non-pip's setting as Python would
# only search the first bigdl package and end up finding only one sub-package.
from .utils import *

View file

@ -19,4 +19,4 @@
# Otherwise there would be module not found error in non-pip's setting as Python would
# only search the first bigdl package and end up finding only one sub-package.
from .log4Error import invalidInputError
from .log4Error import invalidInputError, invalidOperationError

View file

@ -0,0 +1,32 @@
#
# Copyright 2016 The BigDL Authors.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
import sys
from bigdl.llm.utils.common import invalidInputError, invalidOperationError
def get_avx_flags():
avx = ""
if sys.platform != "win32":
import subprocess
msg = subprocess.check_output(["lscpu"]).decode("utf-8")
if "avx512_vnni" in msg:
avx = "_avx512"
elif "avx2" in msg:
avx = "_avx2"
else:
invalidOperationError(False, "Unsupported CPUFLAGS.")
return avx