TurboTransformers：一种快速且用户友好的运行时，用于在CPU和GPU上进行变压器推断（Bert，Albert，GPT2，Decoders等） （274个子文件） 
.clang-format 21B FindMKL.cmake 4KB FindGperftools.cmake 2KB openblas.cmake 1KB eigen.cmake 1KB cuda.cmake 880B transpose.cpp 26KB multi_headed_attention_smart_batch.cpp 19KB transpose_test.cpp 18KB multi_headed_attention.cpp 17KB pybind.cpp 14KB bert_model.cpp 10KB mat_mul.cpp 8KB bert_model_example.cpp 8KB layer_norm.cpp 8KB matmul_benchmark.cpp 7KB bert_model_test.cpp 7KB utils.cpp 6KB seq_pool.cpp 6KB positionwise_ffn.cpp 5KB utils_test.cpp 5KB profiler.cpp 5KB activation.cpp 5KB bert_allocator_test.cpp 5KB bert_config.cpp 4KB allocator_api.cpp 4KB softmax.cpp 4KB mat_mul_test.cpp 4KB model_aware_memory_scheduler.cpp 4KB layernorm_benchmark.cpp 4KB embedding.cpp 4KB tensor.cpp 4KB bert_embedding.cpp 4KB transpose_benchmark.cpp 3KB softmax_test.cpp 3KB albert_layer.cpp 3KB softmax_benchmark.cpp 3KB common.cpp 3KB layer_norm_test.cpp 3KB bert_output.cpp 3KB prepare_bert_masks.cpp 3KB activation_benchmark.cpp 3KB memory.cpp 3KB bert_intermediate.cpp 3KB activation_test.cpp 3KB prepare_bert_masks_test.cpp 3KB embedding_test.cpp 3KB allocator_api_test.cpp 2KB bert_pooler.cpp 2KB gpu_utils_test.cpp 2KB model_aware_memory_scheduler_test.cpp 2KB tensor_test.cpp 2KB blas_openblas.cpp 2KB blas_blis.cpp 2KB bert_attention.cpp 2KB cuda_device_context.cpp 2KB allocator_impl.cpp 2KB config.cpp 1KB enforce.cpp 1KB addbias_layernorm.cpp 1KB enforce_test.cpp 1KB addbias_act.cpp 1KB sequence_pool.cpp 1KB device_context_test.cpp 1KB model_aware_allocator.cpp 967B naive_allocator.cpp 951B base_allocator.cpp 948B fp16_test.cpp 946B ordered_list.cpp 748B benchmark_helper.cpp 702B npz_load.cpp 696B blas_mkl.cpp 692B catch2_test_main.cpp 54B Dockerfile_dev.cpu 1KB Dockerfile_release.cpu 816B gpu_transpose_kernel.cu 16KB gpu_utils.cu 8KB gpu_softmax_kernel.cu 8KB gpu_layer_norm_kernel.cu 5KB gpu_activation_kernel.cu 3KB gpu_embedding_kernel.cu 3KB gpu_block_reduce.cuh 6KB cuda_enforce.cuh 3KB Dockerfile_ci 171B .dockerignore 117B .gitattributes 0B .gitignore 168B .gitmodules 621B Dockerfile_dev.gpu 1KB Dockerfile_release.gpu 1KB tensor.h 10KB multi_headed_attention_smart_batch.h 5KB common.h 5KB model_aware_memory_scheduler.h 5KB multi_headed_attention.h 5KB model_aware_allocator.h 5KB enforce.h 3KB transpose.h 3KB ordered_list.h 3KB gpu_transpose_kernel.h 3KB共 274 条
- 1
- 2
- 3
国服第一奶妈
- 粉丝: 32
- 资源: 4504
最新资源
- 嵌入式开发概述及其常用编程语言介绍
- 5G模组升级刷模块救砖以及5G模组资料路由器固件
- C183579-123578-c1235789.jpg
- Qt5.14 绘画板 Qt Creator C++项目
- python实现Excel表格合并
- Java实现读取Excel批量发送邮件.zip
- 【java毕业设计】商城后台管理系统源码（springboot+vue+mysql+说明文档）.zip
- 【java毕业设计】开发停车位管理系统(调用百度地图API)源码（springboot+vue+mysql+说明文档）.zip
- 星耀软件库(升级版).apk.1
- 基于Django后端和Vue前端的多语言购物车项目设计源码
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
评论0