mirror of
https://github.com/deepspeedai/DeepSpeed.git
synced 2025-10-21 16:48:52 +08:00
f2d4845f07
This PR is to enable quantizer extension on ROCm cc: @jithunnair-amd --------- Co-authored-by: Logan Adams <114770087+loadams@users.noreply.github.com>
39 lines
1.0 KiB
Python
39 lines
1.0 KiB
Python
# Copyright (c) Microsoft Corporation.
|
|
# SPDX-License-Identifier: Apache-2.0
|
|
|
|
# DeepSpeed Team
|
|
|
|
from .builder import CUDAOpBuilder
|
|
|
|
|
|
class QuantizerBuilder(CUDAOpBuilder):
|
|
BUILD_VAR = "DS_BUILD_QUANTIZER"
|
|
NAME = "quantizer"
|
|
|
|
def __init__(self, name=None):
|
|
name = self.NAME if name is None else name
|
|
super().__init__(name=name)
|
|
|
|
def absolute_name(self):
|
|
return f'deepspeed.ops.quantizer.{self.NAME}_op'
|
|
|
|
def sources(self):
|
|
return [
|
|
'csrc/quantization/pt_binding.cpp',
|
|
'csrc/quantization/fake_quantizer.cu',
|
|
'csrc/quantization/quantize.cu',
|
|
'csrc/quantization/quantize_intX.cu',
|
|
'csrc/quantization/dequantize.cu',
|
|
'csrc/quantization/swizzled_quantize.cu',
|
|
'csrc/quantization/quant_reduce.cu',
|
|
]
|
|
|
|
def include_paths(self):
|
|
return ['csrc/includes']
|
|
|
|
def extra_ldflags(self):
|
|
if not self.is_rocm_pytorch():
|
|
return ['-lcurand']
|
|
else:
|
|
return []
|