CMakeLists.txt 8.49 KB
Newer Older
Li Zhang's avatar
Li Zhang committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
# Copyright (c) 2021-2022, NVIDIA CORPORATION. All rights reserved.
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions
# are met:
#  * Redistributions of source code must retain the above copyright
#    notice, this list of conditions and the following disclaimer.
#  * Redistributions in binary form must reproduce the above copyright
#    notice, this list of conditions and the following disclaimer in the
#    documentation and/or other materials provided with the distribution.
#  * Neither the name of NVIDIA CORPORATION nor the names of its
#    contributors may be used to endorse or promote products derived
#    from this software without specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS ``AS IS'' AND ANY
# EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
# PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE COPYRIGHT OWNER OR
# CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
# EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
# PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
# PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY
# OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
# (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
# OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

cmake_minimum_required (VERSION 3.18)

project(tritonfastertransformerbackend LANGUAGES C CXX)

#
# Options
#
option(TRITON_ENABLE_GPU "Enable GPU support in backend" ON)
option(TRITON_ENABLE_STATS "Include statistics collections in backend" ON)

set(TRITON_PYTORCH_INCLUDE_PATHS "" CACHE PATH "Paths to Torch includes")
set(TRITON_PYTORCH_LIB_PATHS "" CACHE PATH "Paths to Torch libraries")

set(TRITON_BACKEND_REPO_TAG "r22.12" CACHE STRING "Tag for triton-inference-server/backend repo")
set(TRITON_CORE_REPO_TAG "r22.12" CACHE STRING "Tag for triton-inference-server/core repo")
set(TRITON_COMMON_REPO_TAG "r22.12" CACHE STRING "Tag for triton-inference-server/common repo")

if(NOT CMAKE_BUILD_TYPE)
  set(CMAKE_BUILD_TYPE Release)
endif()

set(USE_TRITONSERVER_DATATYPE "ON")
message("-- Enable USE_TRITONSERVER_DATATYPE")

#
# Dependencies
#
# FetchContent's composability isn't very good. We must include the
# transitive closure of all repos so that we can override the tag.
#
include(FetchContent)

FetchContent_Declare(
  repo-common
  GIT_REPOSITORY https://github.com/triton-inference-server/common.git
  GIT_TAG ${TRITON_COMMON_REPO_TAG}
  GIT_SHALLOW ON
)
FetchContent_Declare(
  repo-core
  GIT_REPOSITORY https://github.com/triton-inference-server/core.git
  GIT_TAG ${TRITON_CORE_REPO_TAG}
  GIT_SHALLOW ON
)
FetchContent_Declare(
  repo-backend
  GIT_REPOSITORY https://github.com/triton-inference-server/backend.git
  GIT_TAG ${TRITON_BACKEND_REPO_TAG}
  GIT_SHALLOW ON
)
FetchContent_MakeAvailable(repo-common repo-core repo-backend)

#
# CUDA
#
if(${TRITON_ENABLE_GPU})
  find_package(CUDAToolkit REQUIRED)
endif() # TRITON_ENABLE_GPU

#
# Shared library implementing the Triton Backend API
#
configure_file(libtriton_fastertransformer.ldscript libtriton_fastertransformer.ldscript COPYONLY)

add_library(
  triton-fastertransformer-backend SHARED
  libfastertransformer.cc
)

add_library(
  TritonFasterTransformerBackend::triton-fastertransformer-backend ALIAS triton-fastertransformer-backend
)

find_package(CUDAToolkit REQUIRED)
find_package(CUDA 10.1 REQUIRED)
if (${CUDA_VERSION} GREATER_EQUAL 11.0)
  message(STATUS "Add DCUDA11_MODE")
  add_definitions("-DCUDA11_MODE")
endif()

set(CUDA_PATH ${CUDA_TOOLKIT_ROOT_DIR})

target_compile_definitions(triton-fastertransformer-backend
  PUBLIC
  USE_TRITONSERVER_DATATYPE
  BUILD_MULTI_GPU)

target_include_directories(
  triton-fastertransformer-backend
  PRIVATE
  ${CMAKE_CURRENT_SOURCE_DIR}/src
  ${TRITON_PYTORCH_INCLUDE_PATHS}
  ${Python3_INCLUDE_DIRS}
  ${repo-ft_SOURCE_DIR}
  ${repo-ft_SOURCE_DIR}/3rdparty/cutlass/include
  ${repo-core_SOURCE_DIR}/include
  )

target_link_directories(
  triton-fastertransformer-backend
  PRIVATE
  ${CUDA_PATH}/lib64
  )

target_compile_features(triton-fastertransformer-backend PRIVATE cxx_std_14)

target_compile_options(
  triton-fastertransformer-backend PRIVATE
  $<$<OR:$<CXX_COMPILER_ID:Clang>,$<CXX_COMPILER_ID:AppleClang>,$<CXX_COMPILER_ID:GNU>>:
    -Wall -Wextra -Wno-unused-parameter -Wno-type-limits >#-Werror>
)

if(${TRITON_ENABLE_GPU})
  target_compile_definitions(
    triton-fastertransformer-backend
    PRIVATE TRITON_ENABLE_GPU=1
  )
endif() # TRITON_ENABLE_GPU

set_target_properties(
  triton-fastertransformer-backend
  PROPERTIES
    POSITION_INDEPENDENT_CODE ON
    OUTPUT_NAME triton_fastertransformer
    SKIP_BUILD_RPATH TRUE
    BUILD_WITH_INSTALL_RPATH TRUE
    INSTALL_RPATH_USE_LINK_PATH FALSE
    INSTALL_RPATH "$\{ORIGIN\}"
    LINK_DEPENDS ${CMAKE_CURRENT_BINARY_DIR}/libtriton_fastertransformer.ldscript
q.yao's avatar
q.yao committed
156
    LINK_FLAGS "-Wl,--no-as-needed,--version-script ${CMAKE_CURRENT_BINARY_DIR}/libtriton_fastertransformer.ldscript"
Li Zhang's avatar
Li Zhang committed
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
)

# Need to turn off unused-but-set-variable due to Torchvision
# Need to turn off unknown-pragmas due to ATen OpenMP
set_target_properties(
  triton-fastertransformer-backend
  PROPERTIES COMPILE_FLAGS
    "-Wno-unknown-pragmas -Wno-unused-but-set-variable"
)

set(TRITON_PYTORCH_LDFLAGS "")
FOREACH(p ${TRITON_PYTORCH_LIB_PATHS})
  set(TRITON_PYTORCH_LDFLAGS ${TRITON_PYTORCH_LDFLAGS} "-L${p}")
ENDFOREACH(p)

target_link_libraries(
  triton-fastertransformer-backend
  PRIVATE
    triton-core-serverapi  # from repo-core
    triton-core-backendapi # from repo-core
    triton-core-serverstub # from repo-core
    triton-backend-utils   # from repo-backend
    transformer-shared     # from repo-ft
    ${TRITON_PYTORCH_LDFLAGS}
    -lcublas
    -lcublasLt
    -lcudart
    -lcurand
)

if (BUILD_MULTI_GPU)
  target_compile_definitions(
    triton-fastertransformer-backend
    PUBLIC
      BUILD_MULTI_GPU
  )
  target_include_directories(
    triton-fastertransformer-backend
    PRIVATE
      ${MPI_INCLUDE_PATH}
  )
  target_link_directories(
    triton-fastertransformer-backend
    PRIVATE
      ${MPI_Libraries}
      /usr/local/mpi/lib
  )
  target_link_libraries(
    triton-fastertransformer-backend
    PRIVATE
      ${NCCL_LIBRARIES}
      ${MPI_LIBRARIES}
  )
endif()

if(${TRITON_ENABLE_GPU})
  target_link_libraries(
    triton-fastertransformer-backend
    PRIVATE
      CUDA::cudart
  )
endif() # TRITON_ENABLE_GPU

#
# Install
#
include(GNUInstallDirs)
set(INSTALL_CONFIGDIR ${CMAKE_INSTALL_LIBDIR}/cmake/TritonFasterTransformerBackend)

install(
  TARGETS
    triton-fastertransformer-backend
  EXPORT
    triton-fastertransformer-backend-targets
  LIBRARY DESTINATION ${CMAKE_INSTALL_PREFIX}/backends/fastertransformer
  ARCHIVE DESTINATION ${CMAKE_INSTALL_PREFIX}/backends/fastertransformer
)

install(
  EXPORT
    triton-fastertransformer-backend-targets
  FILE
    TritonFasterTransformerBackendTargets.cmake
  NAMESPACE
    TritonFasterTransformerBackend::
  DESTINATION
    ${INSTALL_CONFIGDIR}
)

include(CMakePackageConfigHelpers)
configure_package_config_file(
  ${CMAKE_SOURCE_DIR}/cmake/TritonFasterTransformerBackendConfig.cmake.in
  ${CMAKE_CURRENT_BINARY_DIR}/TritonFasterTransformerBackendConfig.cmake
  INSTALL_DESTINATION ${INSTALL_CONFIGDIR}
)

install(
  FILES
  ${CMAKE_CURRENT_BINARY_DIR}/TritonFasterTransformerBackendConfig.cmake
  DESTINATION ${INSTALL_CONFIGDIR}
)

#
# Export from build tree
#
export(
  EXPORT triton-fastertransformer-backend-targets
  FILE ${CMAKE_CURRENT_BINARY_DIR}/TritonFasterTransformerBackendTargets.cmake
  NAMESPACE TritonFasterTransformerBackend::
)

export(PACKAGE TritonFasterTransformerBackend)


# Copyright (c) 2021-2023, NVIDIA CORPORATION.  All rights reserved.
#
# 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.

add_library(TransformerTritonBackend SHARED transformer_triton_backend.cpp)
target_link_libraries(TransformerTritonBackend PRIVATE nccl_utils mpi_utils)

add_subdirectory(llama)