change FasterTransformer to TurboMind (#37)

CMakeLists.txt

@@ -13,7 +13,7 @@
 # limitations under the License.
 
 cmake_minimum_required(VERSION 3.11 FATAL_ERROR) # for PyTorch extensions, version should be greater than 3.13
-project(FasterTransformer LANGUAGES CXX CUDA)
+project(TurboMind LANGUAGES CXX CUDA)
 
 find_package(CUDA 10.2 REQUIRED)
 
@@ -354,18 +354,18 @@ set_target_properties(transformer-shared PROPERTIES LINKER_LANGUAGE CXX)
 target_link_libraries(transformer-shared PUBLIC -lcudart -lcublas -lcublasLt -lcurand)
 
 include(GNUInstallDirs)
-set(INSTALL_CONFIGDIR ${CMAKE_INSTALL_LIBDIR}/cmake/FasterTransformer)
+set(INSTALL_CONFIGDIR ${CMAKE_INSTALL_LIBDIR}/cmake/TurboMind)
 
 include(CMakePackageConfigHelpers)
 configure_package_config_file(
-  ${CMAKE_CURRENT_LIST_DIR}/cmake/FasterTransformerConfig.cmake.in
-  ${CMAKE_CURRENT_BINARY_DIR}/FasterTransformerConfig.cmake
+  ${CMAKE_CURRENT_LIST_DIR}/cmake/TurboMindConfig.cmake.in
+  ${CMAKE_CURRENT_BINARY_DIR}/TurboMindConfig.cmake
   INSTALL_DESTINATION ${INSTALL_CONFIGDIR}
 )
 
 install(
   FILES
-  ${CMAKE_CURRENT_BINARY_DIR}/FasterTransformerConfig.cmake
+  ${CMAKE_CURRENT_BINARY_DIR}/TurboMindConfig.cmake
   DESTINATION ${INSTALL_CONFIGDIR}
 )
 
@@ -382,7 +382,7 @@ install(
   EXPORT
     transformer-shared-targets
   FILE
-    FasterTransformerTargets.cmake
+    TurboMindTargets.cmake
   DESTINATION
     ${INSTALL_CONFIGDIR}
 )
@@ -391,9 +391,9 @@ export(
   EXPORT
     transformer-shared-targets
   FILE
-    ${CMAKE_CURRENT_BINARY_DIR}/FasterTransformerTargets.cmake
+    ${CMAKE_CURRENT_BINARY_DIR}/TurboMindTargets.cmake
   NAMESPACE
     TritonCore::
 )
 
-export(PACKAGE FasterTransformer)
+export(PACKAGE TurboMind)

cmake/FasterTransformerConfig.cmake.in → cmake/TurboMindConfig.cmake.in

@@ -27,13 +27,13 @@
 include(CMakeFindDependencyMacro)
 
 get_filename_component(
-  FASTERTRANSFORMER_CMAKE_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH
+  TURBOMIND_CMAKE_DIR "${CMAKE_CURRENT_LIST_FILE}" PATH
 )
 
-list(APPEND CMAKE_MODULE_PATH ${FASTERTRANSFORMER_CMAKE_DIR})
+list(APPEND CMAKE_MODULE_PATH ${TURBOMIND_CMAKE_DIR})
 
 if(NOT TARGET transformer-shared)
-  include("${FASTERTRANSFORMER_CMAKE_DIR}/FasterTransformerTargets.cmake")
+  include("${TURBOMIND_CMAKE_DIR}/TurboMindTargets.cmake")
 endif()
 
-set(FASTERTRANSFORMER_LIBRARIES transformer-shared)
+set(TURBOMIND_LIBRARIES transformer-shared)

tests/gemm_dequantize/th_gemm_dequantize.py

@@ -13,14 +13,14 @@ class TestGemmDequantize(unittest.TestCase):
     def setUp(self) -> None:
         torch.classes.load_library('lib/libth_transformer.so')
         torch.classes.load_library('lib/libgemm_dq_unit_ops.so')
-        self.unpack_packed_int4s = torch.ops.fastertransformer.unpack_int4_packed_tensor_to_int8
-        self.pack_int4s = torch.ops.fastertransformer.pack_int8_tensor_to_packed_int4
+        self.unpack_packed_int4s = torch.ops.turbomind.unpack_int4_packed_tensor_to_int8
+        self.pack_int4s = torch.ops.turbomind.pack_int8_tensor_to_packed_int4
         self.fused_gemm_dq = torch.ops.gemm_dq_unit_ops.fused_gemm_dq
         self.fused_gemm_dq_bias_act = torch.ops.gemm_dq_unit_ops.fused_gemm_dq_bias_act
         self.bench = torch.ops.gemm_dq_unit_ops.benchmark_against_cublas_fp
-        self.preprocess_weights_for_mixed_gemm = torch.ops.fastertransformer.preprocess_weights_for_mixed_gemm
+        self.preprocess_weights_for_mixed_gemm = torch.ops.turbomind.preprocess_weights_for_mixed_gemm
 
-        self.symmetric_quantizer = torch.ops.fastertransformer._symmetric_quantize_last_axis_of_batched_matrix
+        self.symmetric_quantizer = torch.ops.turbomind._symmetric_quantize_last_axis_of_batched_matrix
 
         torch.manual_seed(734876213)