Lalithnarayan Cgemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>Luka GovedičSigned-off-by: Lalithnarayan C <Lalithnarayan.C@amd.com> Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
25 lines
782 B
Python
25 lines
782 B
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
|
|
from vllm.logger import init_logger
|
|
from vllm.platforms.cpu import CpuPlatform
|
|
|
|
logger = init_logger(__name__)
|
|
|
|
|
|
class ZenCpuPlatform(CpuPlatform):
|
|
"""CPU platform with AMD Zen (ZenDNN/zentorch) optimizations.
|
|
|
|
Model-load time (dispatch_cpu_unquantized_gemm in layers/utils.py):
|
|
- Routes linear ops to zentorch_linear_unary.
|
|
- When VLLM_ZENTORCH_WEIGHT_PREPACK=1 (default), eagerly prepacks
|
|
weights via zentorch_weight_prepack_for_linear.
|
|
"""
|
|
|
|
device_name: str = "cpu"
|
|
device_type: str = "cpu"
|
|
|
|
def is_zen_cpu(self) -> bool:
|
|
# is_cpu() also returns True for this platform (inherited from CpuPlatform).
|
|
return True
|