Skip to content

Instantly share code, notes, and snippets.

@takiyu

takiyu/dirt_install.txt

Last active May 22, 2019 13:25
Show Gist options
  • Save takiyu/08f339f580806b7878aeddcf33a5f711 to your computer and use it in GitHub Desktop.
Save takiyu/08f339f580806b7878aeddcf33a5f711 to your computer and use it in GitHub Desktop.
Download ZIP
Dirt Install Memo
Raw
dirt_install.txt
$ # Install bazel 0.5.4
$ wget https://github.com/bazelbuild/bazel/releases/download/0.5.4/bazel-0.5.4-installer-linux-x86_64.sh
$ chmod +x bazel-0.5.4-installer-linux-x86_64.sh
$ ./bazel-0.5.4-installer-linux-x86_64.sh --user
$ # Install tensorflow 1.4 (CUDA 10.1, cuDNN 7.5.1)
$ git clone https://github.com/tensorflow/tensorflow.git
$ git checkout r1.4
$ ./configure --disable-multilib
WARNING: ignoring _JAVA_OPTIONS in environment.
WARNING: Running Bazel server needs to be killed, because the startup options are different.
You have bazel 0.5.4 installed.
Please specify the location of python. [Default is /usr/bin/python]: /usr/bin/python2
Found possible Python library paths:
/usr/lib/python2.7/site-packages
/usr/local/lib/python3.7/site-packages
Please input the desired Python library path to use. Default is [/usr/lib/python2.7/site-packages]
Do you wish to build TensorFlow with jemalloc as malloc support? [Y/n]: n
No jemalloc as malloc support will be enabled for TensorFlow.
Do you wish to build TensorFlow with Google Cloud Platform support? [Y/n]: n
No Google Cloud Platform support will be enabled for TensorFlow.
Do you wish to build TensorFlow with Hadoop File System support? [Y/n]: n
No Hadoop File System support will be enabled for TensorFlow.
Do you wish to build TensorFlow with Amazon S3 File System support? [Y/n]: n
No Amazon S3 File System support will be enabled for TensorFlow.
Do you wish to build TensorFlow with XLA JIT support? [y/N]: n
No XLA JIT support will be enabled for TensorFlow.
Do you wish to build TensorFlow with GDR support? [y/N]: n
No GDR support will be enabled for TensorFlow.
Do you wish to build TensorFlow with VERBS support? [y/N]: n
No VERBS support will be enabled for TensorFlow.
Do you wish to build TensorFlow with OpenCL support? [y/N]: n
No OpenCL support will be enabled for TensorFlow.
Do you wish to build TensorFlow with CUDA support? [y/N]: y
CUDA support will be enabled for TensorFlow.
Please specify the CUDA SDK version you want to use, e.g. 7.0. [Leave empty to default to CUDA 8.0]: 10.1
Please specify the location where CUDA 10.1 toolkit is installed. Refer to README.md for more details. [Default is /opt/cuda]:
Please specify the cuDNN version you want to use. [Leave empty to default to cuDNN 6.0]: 7.5.1
Please specify the location where cuDNN 7.5.1 library is installed. Refer to README.md for more details. [Default is /opt/cuda]:
Please specify a list of comma-separated Cuda compute capabilities you want to build with.
You can find the compute capability of your device at: https://developer.nvidia.com/cuda-gpus.
Please note that each additional compute capability significantly increases your build time and binary size. [Default is: 6.1]
Do you want to use clang as CUDA compiler? [y/N]:
nvcc will be used as CUDA compiler.
Please specify which gcc should be used by nvcc as the host compiler. [Default is /usr/bin/gcc]:
Do you wish to build TensorFlow with MPI support? [y/N]:
No MPI support will be enabled for TensorFlow.
Please specify optimization flags to use during compilation when bazel option "--config=opt" is specified [Default is -march=native]:
Add "--config=mkl" to your bazel command to build with MKL support.
Please note that MKL on MacOS or windows is still not supported.
If you would like to use a local MKL instead of downloading, please set the environment variable "TF_MKL_ROOT" every time before build.
Configuration finished
$ # Fix CUDA version mismatch
$ sudo ln -s $CUDA_HOME/include/crt/math_functions.hpp $CUDA_HOME/include/
$ # Add patch for protocol-buffer as following for CUDA>=10.
```
diff --git a/third_party/protobuf/add_noinlines.patch b/third_party/protobuf/add_noinlines.patch
index af74798f06..c1a06ded6d 100644
--- a/third_party/protobuf/add_noinlines.patch
+++ b/third_party/protobuf/add_noinlines.patch
@@ -28,3 +28,39 @@ diff -u -r a/src/google/protobuf/compiler/cpp/cpp_file.cc b/src/google/protobuf/
" static GOOGLE_PROTOBUF_DECLARE_ONCE(once);\n"
" ::google::protobuf::GoogleOnceInit(&once, &AddDescriptorsImpl);\n"
"}\n");
+
+diff --git a/src/google/protobuf/map.h b/src/google/protobuf/map.h
+index 40f35e92..1af91f9c 100644
+--- a/src/google/protobuf/map.h
++++ b/src/google/protobuf/map.h
+@@ -425,10 +425,10 @@ class Map {
+ node_ = NULL;
+ for (bucket_index_ = start_bucket; bucket_index_ < m_->num_buckets_;
+ bucket_index_++) {
+- if (m_->TableEntryIsNonEmptyList(bucket_index_)) {
++ if ((*m_).TableEntryIsNonEmptyList(bucket_index_)) {
+ node_ = static_cast<Node*>(m_->table_[bucket_index_]);
+ break;
+- } else if (m_->TableEntryIsTree(bucket_index_)) {
++ } else if ((*m_).TableEntryIsTree(bucket_index_)) {
+ Tree* tree = static_cast<Tree*>(m_->table_[bucket_index_]);
+ GOOGLE_DCHECK(!tree->empty());
+ node_ = NodePtrFromKeyPtr(*tree->begin());
+@@ -487,7 +487,7 @@ class Map {
+ return true;
+ // Less common: the bucket is a linked list with node_ somewhere in it,
+ // but not at the head.
+- if (m_->TableEntryIsNonEmptyList(bucket_index_)) {
++ if ((*m_).TableEntryIsNonEmptyList(bucket_index_)) {
+ Node* l = static_cast<Node*>(m_->table_[bucket_index_]);
+ while ((l = l->next) != NULL) {
+ if (l == node_) {
+@@ -501,7 +501,7 @@ class Map {
+ // find-like method that compares Node* instead of const Key&.
+ iterator_base i(m_->find(*KeyPtrFromNodePtr(node_), it));
+ bucket_index_ = i.bucket_index_;
+- return m_->TableEntryIsList(bucket_index_);
++ return (*m_).TableEntryIsList(bucket_index_);
+ }
+
+ Node* node_;
```
$ # Fix for cuda compiler compatibility (TODO)
```
diff --git a/tensorflow/core/kernels/reduction_gpu_kernels.cu.h b/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
index be9a611881..cda46570bf 100644
--- a/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
+++ b/tensorflow/core/kernels/reduction_gpu_kernels.cu.h
@@ -268,7 +268,8 @@ __global__ void ColumnReduceMax16ColumnsKernel(
// 1D array necessary due to bug in CUDA 9 compiler.
// TODO(nluehr) revert to 2D array when compiler is ready.
- __shared__ value_type partial_sums[32 * 33];
+ value_type partial_sums[32 * 33];
+ // __shared__ value_type partial_sums[32 * 33];
row += rows_per_warp * gridDim.y * blockDim.y;
for (; row < num_rows; row += rows_per_warp * gridDim.y * blockDim.y) {
@@ -317,7 +318,8 @@ __global__ void ColumnReduceKernel(
// 1D array necessary due to bug in CUDA 9 compiler.
// TODO(nluehr) revert to 2D array when compiler is ready.
- __shared__ value_type partial_sums[32 * 33];
+ value_type partial_sums[32 * 33];
+ // __shared__ value_type partial_sums[32 * 33];
row += gridDim.y * blockDim.y;
```
$ bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package
$ ./bazel-bin/tensorflow/tools/pip_package/build_pip_package $PWD/tensorflow_pkg
$ sudo pip2 install tensorflow_pkg/tensorflow-1.4.2-cp27-cp27mu-linux_x86_64.whl
$ bazel build --config opt --config=cuda --copt=-march=native tensorflow:libtensorflow.so
$ git clone https://github.com/pmh47/dirt.git
$ cd dirt
$ # Change tensorflow path
```
diff --git a/csrc/CMakeLists.txt b/csrc/CMakeLists.txt
index e69ba8d..82305c4 100644
--- a/csrc/CMakeLists.txt
+++ b/csrc/CMakeLists.txt
@@ -12,14 +12,17 @@ find_library(EGL_LIBRARIES NAMES egl EGL REQUIRED)
get_filename_component(NVCC_DIR ${CMAKE_CUDA_COMPILER} DIRECTORY)
find_path(CUDA_INCLUDE_DIR NAMES cuda/include/cuda.h HINTS ${NVCC_DIR}/../.. PATHS ENV CUDA_HOME REQUIRED)
-# Ask tensorflow for its include path; one should therefore make sure cmake is run with the venv active that the op will be used in
-execute_process(COMMAND python -c "import tensorflow; print(tensorflow.sysconfig.get_include())" OUTPUT_VARIABLE Tensorflow_default_INCLUDE_DIRS OUTPUT_STRIP_TRAILING_WHITESPACE)
-set(Tensorflow_INCLUDE_DIRS "${Tensorflow_default_INCLUDE_DIRS}" CACHE PATH "Tensorflow include path")
-
-# Ask tensorflow for its library path
-# If using tensorflow earlier than v1.4, this will not work, but can be skipped entirely
-execute_process(COMMAND python -c "import tensorflow; print(tensorflow.sysconfig.get_lib())" OUTPUT_VARIABLE Tensorflow_default_LIB_DIR OUTPUT_STRIP_TRAILING_WHITESPACE)
-set(Tensorflow_LIB_DIR "${Tensorflow_default_LIB_DIR}" CACHE PATH "Tensorflow library path")
+set(Tensorflow_INCLUDE_DIRS /usr/lib/python2.7/site-packages/tensorflow/include)
+set(Tensorflow_LIB_DIR /home/takiyu/Projects/work/huawei/dirt/tensorflow/bazel-bin/tensorflow/)
+
+# # Ask tensorflow for its include path; one should therefore make sure cmake is run with the venv active that the op will be used in
+# execute_process(COMMAND python -c "import tensorflow; print(tensorflow.sysconfig.get_include())" OUTPUT_VARIABLE Tensorflow_default_INCLUDE_DIRS OUTPUT_STRIP_TRAILING_WHITESPACE)
+# set(Tensorflow_INCLUDE_DIRS "${Tensorflow_default_INCLUDE_DIRS}" CACHE PATH "Tensorflow include path")
+#
+# # Ask tensorflow for its library path
+# # If using tensorflow earlier than v1.4, this will not work, but can be skipped entirely
+# execute_process(COMMAND python2 -c "import tensorflow; print(tensorflow.sysconfig.get_lib())" OUTPUT_VARIABLE Tensorflow_default_LIB_DIR OUTPUT_STRIP_TRAILING_WHITESPACE)
+# set(Tensorflow_LIB_DIR "${Tensorflow_default_LIB_DIR}" CACHE PATH "Tensorflow library path")
find_library(Tensorflow_LIBRARY tensorflow_framework HINTS ${Tensorflow_LIB_DIR} REQUIRED DOC "Tensorflow framework library; for tensorflow < 1.4, you can set this to blank")
# in the following, we need ../external/tensorflow for cuda_config.h in tf versions with #16959 unfixed
diff --git a/setup.py b/setup.py
index 568c5c4..1957645 100644
--- a/setup.py
+++ b/setup.py
@@ -34,7 +34,7 @@ setup(
url='https://github.com/pmh47/dirt',
packages=['dirt'],
python_requires='>=2.7.0',
- install_requires=['tensorflow-gpu>=1.4', 'numpy'],
+ install_requires=[],
package_data={'dirt': ['*.so', '*.dll']},
include_package_data=True,
cmdclass={'build': CmakeAndBuild},
:
ensorflow_default_INCLUDE_DIRS OUTPUT_STRIP_TRAILING_WHITESPACE)
+# set(Tensorflow_INCLUDE_DIRS "${Tensorflow_default_INCLUDE_DIRS}" CACHE PATH "Tensorflow include path")
+#
+# # Ask tensorflow for its library path
+# # If using tensorflow earlier than v1.4, this will not work, but can be skipped entirely
+# execute_process(COMMAND python2 -c "import tensorflow; print(tensorflow.sysconfig.get_lib())" OUTPUT_VARIABLE Tensorflow_default_LIB_DIR OUTPUT_STRIP_TRAILING_WHITESPACE)
+# set(Tensorflow_LIB_DIR "${Tensorflow_default_LIB_DIR}" CACHE PATH "Tensorflow library path")
find_library(Tensorflow_LIBRARY tensorflow_framework HINTS ${Tensorflow_LIB_DIR} REQUIRED DOC "Tensorflow framework library; for tensorflow < 1.4, you can set this to blank")
# in the following, we need ../external/tensorflow for cuda_config.h in tf versions with #16959 unfixed
diff --git a/setup.py b/setup.py
index 568c5c4..1957645 100644
--- a/setup.py
+++ b/setup.py
@@ -34,7 +34,7 @@ setup(
url='https://github.com/pmh47/dirt',
packages=['dirt'],
python_requires='>=2.7.0',
- install_requires=['tensorflow-gpu>=1.4', 'numpy'],
+ install_requires=[],
package_data={'dirt': ['*.so', '*.dll']},
include_package_data=True,
cmdclass={'build': CmakeAndBuild},
~
```
$ pip2 install .
$ python2
Python 2.7.16 (default, Mar 11 2019, 18:59:25)
[GCC 8.2.1 20181127] on linux2
Type "help", "copyright", "credits" or "license" for more information.
>>> import dirt
>>>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment