fblualib在ubuntu 16.04上的安装

topic
4.6k 词

最近需要复现一下CRNN的结果，因为自己实现的版本点数不够高，所以需要跑一下人家的代码。一共有两个依赖，一个是TORCH，这个好装，另一个是fblualib，这个项目已经停止维护了，真的是跳了无数个坑，这里做一个记录。最后也只能跑inference不能train BTW. 但是对于其他跳坑的人可能也是有帮助的.


概览

首先是需要安装的包
fblualib,folly,mstch,zstd,wangle,thpp,fbthrift
其次：
system：ubuntu 16.04，相比来说14.04应该更好装
python:2.7，之前试了python3，但是在import frontend的时候遇到了一些问题，没有解决好，就放弃了，其实是有希望的。

torch

commit 0219027e6c4644a0ba5c5bf137c989a0a8c9e01b

cudnn.torch

1
git clone https://github.com/soumith/cudnn.torch

googletest

1
2
3
4
5
6
7
wget https://github.com/google/googletest/archive/release-1.8.0.tar.gz && 
tar zxf release-1.8.0.tar.gz && 
rm -f release-1.8.0.tar.gz && 
cd googletest-release-1.8.0 && 
cmake configure . && 
make && 
make install

folly

commit ID:5817bad
依赖：

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
sudo apt-get install 
    g++ 
    cmake 
    libboost-all-dev 
    libevent-dev 
    libdouble-conversion-dev 
    libgoogle-glog-dev 
    libgflags-dev 
    libiberty-dev 
    liblz4-dev 
    liblzma-dev 
    libsnappy-dev 
    make 
    zlib1g-dev 
    binutils-dev 
    libjemalloc-dev 
    libssl-dev 
    pkg-config
    libunwind8-dev 
    libelf-dev 
    libdwarf-dev

1
2
3
4
5
6
7
git clone https://github.com/facebook/folly/
cd folly/folly
autoreconf -ivf
./configure &&
make &&
sudo make install &&
sudo ldconfig &&

mstch

commit ID:0fde1cf

1
2
3
4
5
6
7
git clone https://github.com/no1msd/mstch
cd mstch
mkdir build
cd build
cmake ..
make -j 4
sudo make install

zstd

commit ID:87125c2 dev branch

1
2
3
git clone https://github.com/facebook/zstd
cd zstd
sudo make install

googletest

1
2
git clone https://github.com/google/googletest
cmake make install

wangle

commit ID:`31fbaba

1
2
3
4
5
git clone https://github.com/facebook/wangle
cmake .
make
ctest
sudo make install

fbthrift

commit ID:94a399a
更改的部分

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
diff --git a/CMakeLists.txt b/CMakeLists.txt
index 51271e3..7a15eae 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -8,6 +8,9 @@ set(CMAKE_ARCHIVE_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_LIBRARY_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/lib)
 set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${CMAKE_BINARY_DIR}/bin)

+set(CMAKE_CXX_FLAGS "-fPIC")
+set(CMAKE_C_FLAGS "-fPIC")
+
 set(TEMPLATES_INSTALL_DIR include/thrift/templates CACHE STRING
     "The subdirectory where compiler template files should be installed")
 set(BIN_INSTALL_DIR bin CACHE STRING
主要为了解决python包 thrift-compiler无法生成的问题

1
2
3
cd build
cmake ..
make -j 32

thpp

commit ID:0a4c745

1
2
3
git clone https://github.com/facebook/thpp
cd thpp/thpp
./build.sh

更改的部分

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
diff --git a/thpp/CMakeLists.txt b/thpp/CMakeLists.txt
index 4ae3683..603702b 100644
--- a/thpp/CMakeLists.txt
+++ b/thpp/CMakeLists.txt
@@ -42,7 +42,7 @@ ELSE()
   ADD_DEFINITIONS(-DNO_THRIFT)
 ENDIF()

-SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++11")
+SET(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -std=gnu++14")

 SET(src
   Storage.cpp
------------------------------------------------------
   diff --git a/thpp/build.sh b/thpp/build.sh
   index 79af5d9..7c807e7 100755
   --- a/thpp/build.sh
   +++ b/thpp/build.sh
   @@ -52,4 +52,4 @@ make
    ctest

    # Install
   -make install
   +sudo make install
------------------------------------------------------
diff --git a/thpp/detail/TensorGeneric.h b/thpp/detail/TensorGeneric.h
index a8837f6..6b67c72 100644
--- a/thpp/detail/TensorGeneric.h
+++ b/thpp/detail/TensorGeneric.h
@@ -188,17 +188,17 @@ template <> struct TensorOps<Tensor<real>> {
   }
   static void _max(THTensor* values, THLongTensor* indices,
                    THTensor* t, int dim) {
-    return THTensor_(max)(values, indices, t, dim);
+    return THTensor_(max)(values, indices, t, dim, 1);
   }
   static void _min(THTensor* values, THLongTensor* indices,
                    THTensor* t, int dim) {
-    return THTensor_(min)(values, indices, t, dim);
+    return THTensor_(min)(values, indices, t, dim, 1);
   }
   static void _sum(THTensor* r, THTensor* t, int dim) {
-    return THTensor_(sum)(r, t, dim);
+    return THTensor_(sum)(r, t, dim, 1);
   }
   static void _prod(THTensor* r, THTensor* t, int dim) {
-    return THTensor_(prod)(r, t, dim);
+    return THTensor_(prod)(r, t, dim, 1);
   }
   static void _cumsum(THTensor* r, THTensor* t, int dim) {
     return THTensor_(cumsum)(r, t, dim);

fblualib

commit ID:4812779
更改的部分，这个改的太多了

1
2
3
4
5
6
7
8
9
10
11
modified:   fblualib/build.sh
modified:   fblualib/ffivector/CMakeLists.txt
modified:   fblualib/ffivector/FFIVector.cpp
modified:   fblualib/mattorch/CMakeLists.txt
modified:   fblualib/python/CMakeLists.txt
modified:   fblualib/thrift/ChunkedCompression.h
modified:   fblualib/thrift/Encoding.h
modified:   fblualib/thrift/LuaObject.h
modified:   fblualib/thrift/LuaSerialization.cpp
modified:   fblualib/torch/CMakeLists.txt
modified:   fblualib/util/Reactor.cpp

CRNN

ptr的问题

各种报错的解决办法

1
2
3
4
5
6
7
Could not find a package configuration file provided by "Torch" with any of
  the following names:

    TorchConfig.cmake
    torch-config.cmake

export CMAKE_PREFIX_PATH=/path/to/torch