当前位置： 首页 > news >正文

阿里云做网站的代码广东百度seo

news 2025/7/21 0:03:37
阿里云做网站的代码,广东百度seo,运城有做网站设计,商丘疫情写在前面：以下内容完成于2019年底，只是把笔记放到了CSDN上。 需要注释掉NCLL及分布式相关的配置 libcudart.patch diff --git a/torch/cuda/__init__.py b/torch/cuda/__init__.py index 4591702..07e1268 100644 --- a/torch/cuda/__init__.pyb/torc…

写在前面：以下内容完成于2019年底，只是把笔记放到了CSDN上。

需要注释掉NCLL及分布式相关的配置

libcudart.patch

diff --git a/torch/cuda/__init__.py b/torch/cuda/__init__.py
index 4591702..07e1268 100644
--- a/torch/cuda/__init__.py
+++ b/torch/cuda/__init__.py
@@ -59,7 +59,7 @@ def _load_cudart():if platform.system() == 'Windows':lib = find_cuda_windows_lib()else:
-        lib = ctypes.cdll.LoadLibrary(None)
+        lib = ctypes.cdll.LoadLibrary("libcudart.so")if hasattr(lib, 'cudaGetErrorName'):return lib

remove_nccl.patch

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 159b153..6f7423d 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -95,7 +95,7 @@ option(USE_LMDB "Use LMDB" ON)option(USE_METAL "Use Metal for iOS build" ON)option(USE_MOBILE_OPENGL "Use OpenGL for mobile code" ON)option(USE_NATIVE_ARCH "Use -march=native" OFF)
-option(USE_NCCL "Use NCCL" ON)
+option(USE_NCCL "Use NCCL" OFF)option(USE_SYSTEM_NCCL "Use system-wide NCCL" OFF)option(USE_NNAPI "Use NNAPI" OFF)option(USE_NNPACK "Use NNPACK" ON)
@@ -119,7 +119,7 @@ option(USE_TENSORRT "Using Nvidia TensorRT library" OFF)option(USE_ZMQ "Use ZMQ" OFF)option(USE_ZSTD "Use ZSTD" OFF)option(USE_MKLDNN "Use MKLDNN" OFF)
-option(USE_DISTRIBUTED "Use distributed" ON)
+option(USE_DISTRIBUTED "Use distributed" OFF)cmake_dependent_option(USE_MPI "Use MPI for Caffe2. Only available if USE_DISTRIBUTED is on." ON"USE_DISTRIBUTED" OFF)
diff --git a/tools/setup_helpers/dist_check.py b/tools/setup_helpers/dist_check.py
index 8859fe1..5d2ed1c 100644
--- a/tools/setup_helpers/dist_check.py
+++ b/tools/setup_helpers/dist_check.py
@@ -6,7 +6,7 @@ from .env import IS_CONDA, IS_LINUX, IS_WINDOWS, CONDA_DIR, check_env_flag, checfrom .cuda import USE_CUDA# On ROCm, RCCL development isn't complete. https://github.com/ROCmSoftwarePlatform/rccl
-USE_DISTRIBUTED = not check_negative_env_flag("USE_DISTRIBUTED") and not IS_WINDOWS and not check_env_flag("USE_ROCM")
+USE_DISTRIBUTED = FalseUSE_GLOO_IBVERBS = FalseIB_DEVINFO_CMD = "ibv_devinfo"
diff --git a/tools/setup_helpers/nccl.py b/tools/setup_helpers/nccl.py
index c1cc886..576f74e 100644
--- a/tools/setup_helpers/nccl.py
+++ b/tools/setup_helpers/nccl.py
@@ -9,7 +9,7 @@ from .env import IS_WINDOWS, IS_DARWIN, IS_CONDA, CONDA_DIR, check_negative_env_from .cuda import USE_CUDA, CUDA_HOME-USE_NCCL = USE_CUDA and not IS_DARWIN and not IS_WINDOWS
+USE_NCCL = FalseUSE_SYSTEM_NCCL = FalseNCCL_LIB_DIR = NoneNCCL_SYSTEM_LIB = None
diff --git a/torch/CMakeLists.txt b/torch/CMakeLists.txt
index 9c4a018..6849cb1 100644
--- a/torch/CMakeLists.txt
+++ b/torch/CMakeLists.txt
@@ -690,7 +690,7 @@ if (BUILD_PYTHON)list(APPEND TORCH_PYTHON_SRCS ${TORCH_SRC_DIR}/csrc/distributed/c10d/init.cpp)list(APPEND TORCH_PYTHON_LINK_LIBRARIES c10d)list(APPEND TORCH_PYTHON_COMPILE_DEFINITIONS USE_C10D)
-      if (USE_CUDA)
+      if (USE_CUDA AND USE_NCCL)list(APPEND TORCH_PYTHON_SRCS ${TORCH_SRC_DIR}/csrc/distributed/c10d/ddp.cpp)endif()endif()
diff --git a/torch/csrc/distributed/c10d/init.cpp b/torch/csrc/distributed/c10d/init.cpp
index 2b42e1e..11a866d 100644
--- a/torch/csrc/distributed/c10d/init.cpp
+++ b/torch/csrc/distributed/c10d/init.cpp
@@ -435,7 +435,7 @@ They are used in specifying strategies for reduction collectives, e.g.,&::c10d::ProcessGroup::Work::wait,py::call_guard<py::gil_scoped_release>());-#ifdef USE_CUDA
+#if defined(USE_CUDA) && defined(USE_TORCH)module.def("_dist_bucket_tensors",&::c10d::bucketTensors,

iGPU上显示资源不足：iGPU的寄存器数据只有dGPU的一半，需要将降低CUDA的线程数

RuntimeError: cuda runtime error (7) : too many resources requested for launch at /opt/zhangmm/docker/pytorch_1.1.0/aten/src/THCUNN/generic/SpatialUpSamplingBilinear.cu:67

cankao: https://github.com/pytorch/pytorch/issues/8103#issuecomment-424343705

Pytorch on PX2

diff --git a/aten/src/ATen/cuda/CUDAContext.cpp b/aten/src/ATen/cuda/CUDAContext.cpp
index 70a7d05b6..48bf1173e 100644
--- a/aten/src/ATen/cuda/CUDAContext.cpp
+++ b/aten/src/ATen/cuda/CUDAContext.cpp
@@ -24,6 +24,8 @@ void initCUDAContextVectors() {void initDeviceProperty(DeviceIndex device_index) {cudaDeviceProp device_prop;AT_CUDA_CHECK(cudaGetDeviceProperties(&device_prop, device_index));
+  // patch for "too many resources requested for launch"
+  device_prop.maxThreadsPerBlock = device_prop.maxThreadsPerBlock / 2;device_properties[device_index] = device_prop;}diff --git a/aten/src/ATen/cuda/detail/KernelUtils.h b/aten/src/ATen/cuda/detail/KernelUtils.h
index e535f4d83..ac057c504 100644
--- a/aten/src/ATen/cuda/detail/KernelUtils.h
+++ b/aten/src/ATen/cuda/detail/KernelUtils.h
@@ -12,7 +12,10 @@ namespace at { namespace cuda { namespace detail {for (int i = blockIdx.x * blockDim.x + threadIdx.x; i < (n); i += blockDim.x * gridDim.x)// Use 1024 threads per block, which requires cuda sm_2x or above
-constexpr int CUDA_NUM_THREADS = 1024;
+//constexpr int CUDA_NUM_THREADS = 1024;
+
+// patch for "too many resources requested for launch"
+constexpr int CUDA_NUM_THREADS = 512;// CUDA: number of blocks for threads.inline int GET_BLOCKS(const int N)
diff --git a/aten/src/THCUNN/common.h b/aten/src/THCUNN/common.h
index 9e3ed7d85..08fcb4532 100644
--- a/aten/src/THCUNN/common.h
+++ b/aten/src/THCUNN/common.h
@@ -9,7 +9,10 @@"Some of weight/gradient/input tensors are located on different GPUs. Please move them to a single one.")// Use 1024 threads per block, which requires cuda sm_2x or above
-const int CUDA_NUM_THREADS = 1024;
+//const int CUDA_NUM_THREADS = 1024;
+
+// patch for "too many resources requested for launch"
+const int CUDA_NUM_THREADS = 512;// CUDA: number of blocks for threads.inline int GET_BLOCKS(const int N)
http://www.ds6.com.cn/news/44381.html

相关文章：

  • 青岛网站空间seo外链推广工具
  • 网站开发实例社区优化设计答案六年级上册语文
  • app网站开发哪家好百度知道网址
  • 网站banner图最好多大合理百度收录需要多久
  • 淘宝客源码程序 爱淘宝风格+程序自动采集商品 淘宝客网站模板网站开发软件
  • 南通六建网站电商怎么推广自己的产品
  • 响应式网站设计图网络服务包括哪些内容
  • 邮编域名做网站收录批量查询
  • 俄文网站推广获客渠道有哪些
  • 室内设计网站排行榜前十名有哪些站长之家字体
  • 优设网设计服务平台成都网站关键词推广优化
  • 刚做的网站怎么在百度上能搜到西安网站seo哪家公司好
  • 外贸购物网站制作怎么推广
  • wordpress安装好了怎么登陆网站谷歌优化方法
  • wordpress 拿shellseo搜索引擎优化内容
  • 网站框架指的是什么uv推广平台
  • 免费网站模板 带后台seo服务如何收费
  • 新疆网站设计百度指数
  • 专业做国际网站的公司宣传软文范例
  • 悬赏做logo的网站友链网
  • 江西做网站优化好的网上推广的平台有哪些
  • 简单大气的企业网站优化推广关键词
  • 肇庆cms建站系统如何做seo优化
  • 在什么网站做公务员题目模拟搜索点击软件
  • 沧州网络推广杭州专业seo公司
  • 公司网站备案需要什么网址收录大全
  • 中企动力做网站靠谱吗设计师必备的6个网站
  • 建设银行官网站查询河北seo关键词排名优化
  • 宿州做网站公司杭州seo按天计费
  • 网站框架设计好后怎么做泉州seo按天计费