Python3 --data=data/fruit --model-dir=models/fruit --batch-size=4 --epochs=30

/home/ljx/.local/lib/python3.6/site-packages/torch/nn/ UserWarning: size_average and reduce args will be deprecated, please use reduction=‘sum’ instead.
THCudaCheck FAIL file=/local/jetbot/pytorch/aten/src/THCUNN/generic/ line=202 error=701 : too many resources requested for launch
Traceback (most recent call last):
File “”, line 342, in
device=DEVICE, debug_steps=args.debug_steps, epoch=epoch)
File “”, line 123, in train
File “/home/ljx/.local/lib/python3.6/site-packages/torch/”, line 185, in backward
torch.autograd.backward(self, gradient, retain_graph, create_graph)
File “/home/ljx/.local/lib/python3.6/site-packages/torch/autograd/”, line 127, in backward
allow_unreachable=True) # allow_unreachable flag
RuntimeError: cuda runtime error (701) : too many resources requested for launch at /local/jetbot/pytorch/aten/src/THCUNN/generic/



Which PyTorch package do you install?
If you are not using our prebuilt package, please give it a try.

You can find the installation information in the below topic:






Which JetPack version do you use?
If you are using JetPack4.6, would you mind testing PyTorch v1.8.0 with torchvision v0.9.0?




Sorry for the late update.

Please noted that you will also need to use the corresponding jetson_inference branch based on the JetPack version.
Have you checkout with -b L4T-R32.4.4?


wget -O torch-1.9.0-cp36-cp36m-linux_aarch64.whl

Please provide this package and send others. Chinese browsers can’t open and download it

Hi @junxing.liang, here is a mirror on Google Drive to the PyTorch 1.9 wheel:

Here are the other PyTorch wheels for JetPack >= 4.4

您好,这个问题解决了吗?我也遇到了同样的错误,pytorch 1.6 jetpack 4.5

I got the same problem,pytorch 1.6 jetpack 4.5,what can i do to make this work?


非常感谢,我在按照官网安装pytorch1.7时,安装成功,但是在安装torchvision0.8.1时,卡在了一个位置不动了,也没有报错,在torchvision文件夹执行的这个命令python3 install --user卡在了:building ‘torchvision.image’ extension
Emitting ninja build file /home/kavin/torchvision/build/temp.linux-aarch64-3.6/…
Compiling objects…
Allowing ninja to set a default number of workers… (overridable by setting the environment variable MAX_JOBS=N)

sudo apt-get install libjpeg-dev zlib1g-dev libpython3-dev libavcodec-dev libavformat-dev libswscale-dev
$ git clone --branch GitHub - pytorch/vision: Datasets, Transforms and Models specific to Computer Vision torchvision # see below for version of torchvision to download
$ cd torchvision
$ export BUILD_VERSION=0.x.0 # where 0.x.0 is the torchvision version
$ python3 install --user
$ cd …/ # attempting to load torchvision from build dir will result in import error
$ pip install ‘pillow<7’ # always needed for Python 2.7, not needed torchvision v0.5.0+ with Python 3.6