Jiangyun ZhuandLuka Govedič
8060bb0333
Signed-off-by: zjy0516 <riverclouds.zhu@qq.com> Signed-off-by: Jiangyun Zhu <riverclouds.zhu@qq.com> Co-authored-by: Luka Govedič <ProExpertProg@users.noreply.github.com>
34 lines
1.1 KiB
Python
34 lines
1.1 KiB
Python
# SPDX-License-Identifier: Apache-2.0
|
|
# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
|
|
import torch
|
|
from torch import Tensor
|
|
|
|
from vllm import ir
|
|
from vllm.platforms import current_platform
|
|
|
|
current_platform.import_kernels()
|
|
|
|
CUDA_ALIKE = current_platform.is_cuda_alike()
|
|
"""Most kernels in this file are supported on all CUDA-alike platforms."""
|
|
|
|
rms_no_var_size = (
|
|
lambda x, weight, epsilon, variance_size=None: variance_size is None
|
|
and (weight is None or weight.dtype == x.dtype)
|
|
)
|
|
"""vLLM kernel requires no variance_size override and matching input/weight dtype."""
|
|
|
|
|
|
@ir.ops.rms_norm.register_impl(
|
|
"vllm_c", supports_args=rms_no_var_size, supported=CUDA_ALIKE
|
|
)
|
|
def rms_norm(
|
|
x: Tensor, weight: Tensor | None, epsilon: float, variance_size: int | None = None
|
|
) -> Tensor:
|
|
if weight is None:
|
|
# Kernel requires weight tensor, pass ones
|
|
weight = torch.ones(x.shape[-1], device=x.device, dtype=x.dtype)
|
|
assert variance_size is None
|
|
output = torch.empty(x.shape, device=x.device, dtype=x.dtype)
|
|
torch.ops._C.rms_norm(output, x, weight, epsilon)
|
|
return output
|