Install procedure for pyTorch on NVIDIA Jetson TX1/TX2 with JetPack <= 3.2.1. For JetPack 4.2 and Xavier/Nano/TX2, see https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano/

pytorch_jetson_install.sh

#!/bin/bash
#
# EDIT:  this script is outdated, please see https://forums.developer.nvidia.com/t/pytorch-for-jetson-nano-version-1-6-0-now-available
#
sudo apt-get install python-pip

# upgrade pip
pip install -U pip
pip --version
# pip 9.0.1 from /home/ubuntu/.local/lib/python2.7/site-packages (python 2.7)

# clone pyTorch repo
git clone http://github.com/pytorch/pytorch
cd pytorch
git submodule update --init

# install prereqs
sudo pip install -U setuptools
sudo pip install -r requirements.txt

# Develop Mode:
python setup.py build_deps
sudo python setup.py develop

# Install Mode:  (substitute for Develop Mode commands)
#sudo python setup.py install

# Verify CUDA (from python interactive terminal)
# import torch
# print(torch.__version__)
# print(torch.cuda.is_available())
# a = torch.cuda.FloatTensor(2)
# print(a)
# b = torch.randn(2).cuda()
# print(b)
# c = a + b
# print(c)

Its recommended to upgrade to JP 4.2 and use the PyTorch v.1.1.0 or v.1.0.0 pre-compiled pip wheels from the forums: https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano-with-new-torch2trt-converter/

…

________________________________ From: Saravanan <[email protected]> Sent: Tuesday, July 16, 2019 11:31:16 PM To: dusty-nv Cc: Dustin Franklin; Author Subject: Re: dusty-nv/pytorch_jetson_install.sh Hi, Sara here, On Jetson TX2, i followed the steps listed in the script, when i try to build_deps, getting an error. pasted below. Inputs will be good, thanks. python setup.py build_deps Building wheel torch-1.2.0a0+ec57d92 usage: setup.py [global_opts] cmd1 [cmd1_opts] [cmd2 [cmd2_opts] ...] or: setup.py --help [cmd1 cmd2 ...] or: setup.py --help-commands or: setup.py cmd --help error: invalid command 'build_deps' — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://gist.github.com/ef2b372301c00c0a9d3203e42fd83426?email_source=notifications&email_token=ADVEGK56TZMFX6G3D2OF6J3P72HAJA5CNFSM4HJ5YXH2YY3PNVWWK3TUL52HS4DFVNDWS43UINXW23LFNZ2KUY3PNVWWK3TUL5UWJTQAFVNRI#gistcomment-2972436>, or mute the thread<https://github.com/notifications/unsubscribe-auth/ADVEGK44S4DUKE6WP5CSAPTP72HAJANCNFSM4HJ5YXHQ>.

----------------------------------------------------------------------------------- This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. -----------------------------------------------------------------------------------

Hi,
I am trying to install pytorch v1 on TX2 with Jetpack 3.3 and I m getting the following error after running the command $ python setup.py install:
....
subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--target', 'install', '--config', 'Release', '--', '-j', '6']' returned non-zero exit status 2
I also tried to run python setup.py build_deps but I am getting error: invalid command 'build_deps'

Do you know how could I solve it?

Thanks a lot!

I found the following comment in setup.py and it solved my issue (i had the same)

It is no longer necessary to use the 'build' or 'rebuild' targets

To install:
  $ python setup.py install
To develop locally:
  $ python setup.py develop
To force cmake to re-generate native build files (off by default):
  $ python setup.py develop --cmake

See the build instructions from this post for PyTorch v1 -- https://devtalk.nvidia.com/default/topic/1049071/jetson-nano/pytorch-for-jetson-nano-version-1-3-0-now-available/ Note that it disables NCCL through and environment variable before you run setup.py

…

________________________________ From: Attila Door <[email protected]> Sent: Friday, November 22, 2019 5:52:43 AM To: dusty-nv <[email protected]> Cc: Dustin Franklin <[email protected]>; Author <[email protected]> Subject: Re: dusty-nv/pytorch_jetson_install.sh Hi, I am trying to install pytorch v1 on TX2 with Jetpack 3.3 and I m getting the following error after running the command $ python setup.py install: Makefile:68: recipe for target '/home/nvidia/Desktop/pytorch/build/nccl/obj/collectives/device/devlink.o' failed make[5]: *** [/home/nvidia/Desktop/pytorch/build/nccl/obj/collectives/device/devlink.o] Error 255 Makefile:44: recipe for target '/home/nvidia/Desktop/pytorch/build/nccl/obj/collectives/device/colldevice.a' failed make[4]: *** [/home/nvidia/Desktop/pytorch/build/nccl/obj/collectives/device/colldevice.a] Error 2 Makefile:25: recipe for target 'src.build' failed make[3]: *** [src.build] Error 2 CMakeFiles/nccl_external.dir/build.make:110: recipe for target 'nccl_external-prefix/src/nccl_external-stamp/nccl_external-build' failed make[2]: *** [nccl_external-prefix/src/nccl_external-stamp/nccl_external-build] Error 2 CMakeFiles/Makefile2:67: recipe for target 'CMakeFiles/nccl_external.dir/all' failed make[1]: *** [CMakeFiles/nccl_external.dir/all] Error 2 Makefile:138: recipe for target 'all' failed make: *** [all] Error 2 Traceback (most recent call last): File "setup.py", line 749, in <module> build_deps() File "setup.py", line 323, in build_deps cmake=cmake) File "/home/nvidia/Desktop/pytorch/tools/build_pytorch_libs.py", line 64, in build_caffe2 cmake.build(my_env) File "/home/nvidia/Desktop/pytorch/tools/setup_helpers/cmake.py", line 345, in build self.run(build_args, my_env) File "/home/nvidia/Desktop/pytorch/tools/setup_helpers/cmake.py", line 107, in run check_call(command, cwd=self.build_dir, env=env) File "/usr/lib/python3.5/subprocess.py", line 581, in check_call raise CalledProcessError(retcode, cmd) subprocess.CalledProcessError: Command '['cmake', '--build', '.', '--target', 'install', '--config', 'Release', '--', '-j', '6']' returned non-zero exit status 2 I also tried to run python setup.py build_deps but I am getting error: invalid command 'build_deps' Do you know how could I solve it? Thanks a lot! I found the following comment in setup.py and it solved my issue (i had the same) It is no longer necessary to use the 'build' or 'rebuild' targets To install: $ python setup.py install To develop locally: $ python setup.py develop To force cmake to re-generate native build files (off by default): $ python setup.py develop --cmake — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub<https://gist.github.com/ef2b372301c00c0a9d3203e42fd83426?email_source=notifications&email_token=ADVEGK6XE6F57ML37FKNC6LQU62XXA5CNFSM4HJ5YXH2YY3PNVWWK3TUL52HS4DFVNDWS43UINXW23LFNZ2KUY3PNVWWK3TUL5UWJTQAF4TWG#gistcomment-3090275>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ADVEGK6KFTNBGNYIO3UT6ETQU62XXANCNFSM4HJ5YXHQ>.

----------------------------------------------------------------------------------- This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. -----------------------------------------------------------------------------------

Hey @dusty-nv , it seems that the latest release of NCCL 2.6.4.1 recognizes ARM CPUs. I'm currently attempting to install it to my Jetson TX2, because I have been wanting this for some time. However, I must warn: some scripts from the master branch of nccl git are commited with messages from previous releases, which is a yellow flag. If I do get it, I intend to release all my configuration and installation scripts for the community. I'll let you know.

Have a good one.

git submodule update --init --recursive

Hi Everyone, I am having a similar issue and need some assistance. I am trying to install pytorch on the odroid board xu4 but I get the following error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’. Below is the error message directly from the terminal.
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c: In function ‘xnn_qs8_gemm_minmax_fp32_ukernel_1x8c4__neondot’:
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:91:17: error: redefinition of ‘vproduct0x0123’
91 | float32x4_t vproduct0x0123 = vmulq_f32(vproduct0x0123, vscale);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:88:17: note: previous definition of ‘vproduct0x0123’ was here
88 | float32x4_t vproduct0x0123 = vcvtq_f32_s32(vacc0x0123);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:92:17: error: redefinition of ‘vproduct0x4567’
92 | float32x4_t vproduct0x4567 = vmulq_f32(vproduct0x4567, vscale);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:89:17: note: previous definition of ‘vproduct0x4567’ was here
89 | float32x4_t vproduct0x4567 = vcvtq_f32_s32(vacc0x4567);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:94:18: warning: implicit declaration of function ‘vcvtnq_s32_f32’; did you mean ‘vcvtq_s32_f32’? [-Wimplicit-function-declaration]
94 | vacc0x0123 = vcvtnq_s32_f32(vproduct0x0123);
| ^~~~~~~~~~~~~~
| vcvtq_s32_f32
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:94:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c:95:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
95 | vacc0x4567 = vcvtnq_s32_f32(vproduct0x4567);
| ^~~~~~~~~~~~~~
[ 60%] Building C object confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c.o
make[2]: *** [confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/build.make:15598: confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/1x8c4-minmax-fp32-neondot.c.o] Error 1
make[2]: *** Waiting for unfinished jobs....
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c: In function ‘xnn_qs8_gemm_minmax_fp32_ukernel_1x16c4__neondot’:
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c:112:18: warning: implicit declaration of function ‘vcvtnq_s32_f32’; did you mean ‘vcvtq_s32_f32’? [-Wimplicit-function-declaration]
112 | vacc0x0123 = vcvtnq_s32_f32(vproduct0x0123);
| ^~~~~~~~~~~~~~
| vcvtq_s32_f32
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c:112:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c:113:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
113 | vacc0x4567 = vcvtnq_s32_f32(vproduct0x4567);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c:114:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
114 | vacc0x89AB = vcvtnq_s32_f32(vproduct0x89AB);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c:115:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
115 | vacc0xCDEF = vcvtnq_s32_f32(vproduct0xCDEF);
| ^~~~~~~~~~~~~~
make[2]: *** [confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/build.make:15624: confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/1x16c4-minmax-fp32-neondot.c.o] Error 1
/tmp/cc3aAMfl.s: Assembler messages:
/tmp/cc3aAMfl.s:73: Error: selected processor does not support vsdot.s8 q9,q10,d7[0]' in ARM mode /tmp/cc3aAMfl.s:76: Error: selected processor does not support vsdot.s8 q9,q10,d7[1]' in ARM mode
/tmp/cc3aAMfl.s:78: Error: selected processor does not support vsdot.s8 q8,q10,d7[0]' in ARM mode /tmp/cc3aAMfl.s:80: Error: selected processor does not support vsdot.s8 q8,q10,d7[1]' in ARM mode
/tmp/cc3aAMfl.s:152: Error: selected processor does not support vsdot.s8 q9,q10,d7[0]' in ARM mode /tmp/cc3aAMfl.s:154: Error: selected processor does not support vsdot.s8 q8,q10,d7[0]' in ARM mode
make[2]: *** [confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/build.make:15611: confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/1x8c4-minmax-gemmlowp-neondot.c.o] Error 1
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c: In function ‘xnn_qs8_gemm_minmax_fp32_ukernel_4x8c4__neondot’:
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:154:18: warning: implicit declaration of function ‘vcvtnq_s32_f32’; did you mean ‘vcvtq_s32_f32’? [-Wimplicit-function-declaration]
154 | vacc0x0123 = vcvtnq_s32_f32(vproduct0x0123);
| ^~~~~~~~~~~~~~
| vcvtq_s32_f32
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:154:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:155:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
155 | vacc0x4567 = vcvtnq_s32_f32(vproduct0x4567);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:156:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
156 | vacc1x0123 = vcvtnq_s32_f32(vproduct1x0123);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:157:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
157 | vacc1x4567 = vcvtnq_s32_f32(vproduct1x4567);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:158:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
158 | vacc2x0123 = vcvtnq_s32_f32(vproduct2x0123);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:159:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
159 | vacc2x4567 = vcvtnq_s32_f32(vproduct2x4567);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:160:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
160 | vacc3x0123 = vcvtnq_s32_f32(vproduct3x0123);
| ^~~~~~~~~~~~~~
/home/odroid/pytorch_install/pytorch/third_party/XNNPACK/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c:161:18: error: incompatible types when assigning to type ‘int32x4_t’ from type ‘int’
161 | vacc3x4567 = vcvtnq_s32_f32(vproduct3x4567);
| ^~~~~~~~~~~~~~
make[2]: *** [confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/build.make:15650: confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/4x8c4-minmax-fp32-neondot.c.o] Error 1
/tmp/ccB1jg1m.s: Assembler messages:
/tmp/ccB1jg1m.s:80: Error: selected processor does not support vsdot.s8 q12,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:83: Error: selected processor does not support vsdot.s8 q12,q9,d7[1]' in ARM mode
/tmp/ccB1jg1m.s:86: Error: selected processor does not support vsdot.s8 q11,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:89: Error: selected processor does not support vsdot.s8 q11,q9,d7[1]' in ARM mode
/tmp/ccB1jg1m.s:92: Error: selected processor does not support vsdot.s8 q10,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:94: Error: selected processor does not support vsdot.s8 q10,q9,d7[1]' in ARM mode
/tmp/ccB1jg1m.s:96: Error: selected processor does not support vsdot.s8 q8,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:98: Error: selected processor does not support vsdot.s8 q8,q9,d7[1]' in ARM mode
/tmp/ccB1jg1m.s:196: Error: selected processor does not support vsdot.s8 q12,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:198: Error: selected processor does not support vsdot.s8 q10,q9,d7[0]' in ARM mode
/tmp/ccB1jg1m.s:200: Error: selected processor does not support vsdot.s8 q11,q9,d7[0]' in ARM mode /tmp/ccB1jg1m.s:202: Error: selected processor does not support vsdot.s8 q8,q9,d7[0]' in ARM mode
make[2]: *** [confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/build.make:15637: confu-deps/XNNPACK/CMakeFiles/XNNPACK.dir/src/qs8-gemm/gen/1x16c4-minmax-gemmlowp-neondot.c.o] Error 1
make[1]: *** [CMakeFil