cannot run tutorial files well.

[Arthur-Shi]

Python 3.6.1 |Anaconda 4.4.0 (64-bit)| (default, May 11 2017, 13:25:24) [MSC v.1900 64 bit (AMD64)]
Type "copyright", "credits" or "license" for more information.

IPython 5.3.0 -- An enhanced Interactive Python.
? -> Introduction and overview of IPython's features.
%quickref -> Quick reference.
help -> Python's own help system.
object? -> Details about 'object', use 'object??' for extra details.

runfile('C:/Users/admin/tvm/tutorials/get_started.py', wdir='C:/Users/admin/tvm/tutorials')
<class 'tvm.tensor.Tensor'>
Traceback (most recent call last):

File "", line 1, in
runfile('C:/Users/admin/tvm/tutorials/get_started.py', wdir='C:/Users/admin/tvm/tutorials')

File "C:\Program Files\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", line 880, in runfile
execfile(filename, namespace)

File "C:\Program Files\Anaconda3\lib\site-packages\spyder\utils\site\sitecustomize.py", line 102, in execfile
exec(compile(f.read(), filename, 'exec'), namespace)

File "C:/Users/admin/tvm/tutorials/get_started.py", line 111, in
fadd_cuda = tvm.build(s, [A, B, C], "cuda", target_host="llvm", name="myadd")

File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm\build_module.py", line 368, in build
mhost = codegen.build_module(fhost, target_host)

File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm\codegen.py", line 20, in build_module
return _Build(lowered_func, target)

File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi\function.py", line 255, in my_api_func
return flocal(*args)

File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi_ctypes\function.py", line 183, in call
ctypes.byref(ret_val), ctypes.byref(ret_tcode)))

File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi\base.py", line 62, in check_call
raise TVMError(py_str(_LIB.TVMGetLastError()))

TVMError: [14:08:51] C:\Users\admin\tvm\src\codegen\codegen.cc:27: Check failed: bf != nullptr Target llvm is not enabled

I tried other softwares, but it gave the same output.

What should I do to make the file work?

[Arthur-Shi]

Is it not a file to run? Just for introduction of some important functions？

[Arthur-Shi]

sometimes It can output right, but it will go wrong after running again.

[tqchen]

as the error indicates, you will need to build with llvm enabled

[Arthur-Shi]

yeah, I found the llvm folder and which file should be bulit by VS?

Thank you so much.

[tqchen]

@sxjscience do you have some instruction for building with LLVM on windows?

[Arthur-Shi]

And how to install the topi package? there is no setup.py in the topi folder. I tried to find the package.
sys.path.append('\Users\admin\tvm\topi\python\topi')
But it is useless.

It is shown in the console.
File "C:/Users/admin/tvm/topi/recipe/conv/depthwise_conv2d_test.py", line 7, in
import topi

ModuleNotFoundError: No module named 'topi

[sxjscience]

@Arthur-Shi Try to use sys.path.append('C:\\Users\\admin\\tvm\\topi\\python').

Also, you can manually update the easy-install.pth file by adding C:\\Users\\admin\\tvm\\topi\\python. For example, I'm using Anaconda3 and my easy-install.pth file is located in C:\\Anaconda3\\Lib\\site-packages\\easy-install.pth.

For the llvm, I haven't installed it in Windows. However, I feel that it is not so hard by following the document https://llvm.org/docs/GettingStartedVS.html .

[Arthur-Shi]

I am trying to run tvm on Linux, but I still don't know how to build the library?

I tried this method, but I don't know how to programming in the terminal.
http://tldp.org/HOWTO/Program-Library-HOWTO/shared-libraries.html#AEN115

Could you show me the code tips?
cmake -G "Visual Studio 14 2015 Win64" -DCMAKE_BUILD_TYPE=Release -DCMAKE_CONFIGURATION_TYPES="Release" .. I think it is wrong, but how to change? I downloaded G++.

is it installing llvm necessary for linux?
I downloaded the llvm.tar and untar it. and then I added the deb.

[Arthur-Shi]

@sxjscience I tried two methods you said, but still cannot import topi

[sxjscience]

@Arthur-Shi Could you run this?

'''python
import sys
sys.append('C:\Users\admin\tvm\topi\python')
import tvm
'''
If not, what's the error message?

[Arthur-Shi]

@sxjscience It works.

I didnot run import sys

sorry.

Do you know what can i do?

runfile('C:/Users/admin/tvm/topi/recipe/conv/depthwise_conv2d_test.py', wdir='C:/Users/admin/tvm/topi/recipe/conv')
Skip because cuda is not enabled
Skip because cuda is not enabled

I have installed CUDA
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v8.0

[sxjscience]

Try the command here https://github.com/sxjscience/tvm/blob/master/windows_cmake_command.txt

[Arthur-Shi]

It is the output I run this command.

CMake Error: The source directory "C:/Users/admin" does not appear to contain CMakeLists.txt.
Specify --help for usage, or press the help button on the CMake GUI.

[sxjscience]

Please read the doc https://github.com/dmlc/tvm/blob/master/docs/how_to/install.md

[Arthur-Shi]

sorry @sxjscience

Ｉ　ｄｉｄｎｏｔ　ｒｕｎ　ｔｈｉｓ　ｃｏｍｍａｎｄ　ｉｎ　ｔｈｅ　ｂｕｉｌｄ　ｆｏｌｄｅｒ．

C:\Users\admin\tvm\build>cmake -G "Visual Studio 15 2017 Win64" ^
More? -DUSE_CUDA="ON" ^
More? -DCMAKE_BUILD_TYPE=Release ^
More? -DCMAKE_CONFIGURATION_TYPES="Release" ^
More? -DUSE_RPC="ON" ^
More? ..
-- Found CUDA: C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0 (found version "8.0")
-- Build with CUDA support
-- Build with RPC support...
-- Build with Graph runtime support...
-- Configuring done
-- Generating done
-- Build files have been written to: C:/Users/admin/tvm/build

C:\Users\admin\tvm\build>

Ｂｕｔ　
runfile('C:/Users/admin/tvm/topi/recipe/conv/depthwise_conv2d_test.py', wdir='C:/Users/admin/tvm/topi/recipe/conv')
Skip because cuda is not enabled
Skip because cuda is not enabled

[sxjscience]

Please read the doc more carefully. The doc has stated that "This will generate the VS project using the MSVC 14 64 bit generator. Open the .sln file in the build directory and build with Visual Studio.". You will see a .sln file in the "build" directory. You need to open it with visual studio and build the solution file.

[Arthur-Shi]

ye，I built it by MSVC 15.

I dont know why? Some times My output is
runfile('C:/Users/admin/tvm/topi/recipe/conv/depthwise_conv2d_test.py', wdir='C:/Users/admin/tvm/topi/recipe/conv')
Skip because cuda is not enabled
Skip because cuda is not enabled

sometimes my output is

Traceback (most recent call last):
File "C:\Users\admin\tvm\tutorials\get_started.py", line 111, in
fadd_cuda = tvm.build(s, [A, B, C], "cuda", target_host="llvm", name="myadd")
File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm\build_module.py", line 368, in build
mhost = codegen.build_module(fhost, target_host)
File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm\codegen.py", line 20, in build_module
return _Build(lowered_func, target)
File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi\function.py", line 255, in my_api_func
return flocal(*args)
File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi_ctypes\function.py", line 183, in call
ctypes.byref(ret_val), ctypes.byref(ret_tcode)))
File "C:\Users\admin\AppData\Roaming\Python\Python36\site-packages\tvm-0.1.0-py3.6-win-amd64.egg\tvm_ffi\base.py", line 62, in check_call
raise TVMError(py_str(_LIB.TVMGetLastError()))
tvm._ffi.base.TVMError: [16:28:15] C:\Users\admin\tvm\src\codegen\codegen.cc:27: Check failed: bf != nullptr Target llvm is not enabled

Acturally I built the tvm.sln and built the llvm-5.0.0.src. and run ur command.

[Arthur-Shi]

Because I did not switch cuda on in the config.mk. And the I can run on the Ubtuntu.
Thank you so much.

[Arthur-Shi]

#754 continue trying to deal with this problem.

[jsjs0827]

@Arthur-Shi I encoutner the same issue in WINDOWS10;I make the tvm with pre-build llvm(clang already can be used) successfully,can you give me some advice,thanks for any reply.