transformer.py

"""
Copyright 2020 The Microsoft DeepSpeed Team
"""
import torch
from .builder import CUDAOpBuilder


class TransformerBuilder(CUDAOpBuilder):
    BUILD_VAR = "DS_BUILD_TRANSFORMER"
    NAME = "transformer"

    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.transformer.{self.NAME}_op'

    def sources(self, is_rocm_pytorch):
        if is_rocm_pytorch:
            return [
                'csrc/transformer/hip/ds_transformer_hip.cpp',
                'csrc/transformer/hip/cublas_wrappers.hip',
                'csrc/transformer/hip/transform_kernels.hip',
                'csrc/transformer/hip/gelu_kernels.hip',
                'csrc/transformer/hip/dropout_kernels.hip',
         ###don't support
                #'csrc/transformer/hip/normalize_kernels.hip',
                #'csrc/transformer/hip/softmax_kernels.hip',
                'csrc/transformer/hip/general_kernels.hip'
            ]
        else:
            return [
                'csrc/transformer/ds_transformer_cuda.cpp',
                'csrc/transformer/cublas_wrappers.cu',
                'csrc/transformer/transform_kernels.cu',
                'csrc/transformer/gelu_kernels.cu',
                'csrc/transformer/dropout_kernels.cu',
                'csrc/transformer/normalize_kernels.cu',
                'csrc/transformer/softmax_kernels.cu',
                'csrc/transformer/general_kernels.cu'
            ]

    def include_paths(self):
        return ['csrc/includes']

    def nvcc_args(self):
        args = [
            '-O3',
            #'--use_fast_math',
            '-std=c++14',
            #'-U__CUDA_NO_HALF_OPERATORS__',
            #'-U__CUDA_NO_HALF_CONVERSIONS__',
            #'-U__CUDA_NO_HALF2_OPERATORS__'
        ]

        return args #+ self.compute_capability_args()

    def cxx_args(self):
        return ['-O3', '-std=c++14', '-g', '-Wno-reorder', '-Wno-c++11-narrowing']