What is the release schedule for the official Docker pytorch/pytorch:latest distribution?
I did a docker pull today but am still running 0.2.0.
I don’t know about the release schedule, but you can download the GitHub repo and build a docker image (instructions here)
After building the most recent Docker image for PyTorch, and then launching it with nvidia-docker 2.0:
$ docker build -t pytorch_cuda9 -f tools/docker/Dockerfile9 .
Sending build context to Docker daemon 33.78MB
Step 1/14 : FROM nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04
9.0-cudnn7-devel-ubuntu16.04: Pulling from nvidia/cuda
50aff78429b1: Pull complete
f6d82e297bce: Pull complete
275abb2c8a6f: Pull complete
9f15a39356d6: Pull complete
fc0342a94c89: Pull complete
a4adf2161c6b: Pull complete
8f86feaebedb: Pull complete
a1f2e681a892: Pull complete
dddbbfcd9343: Pull complete
e16144f996cb: Pull complete
8cd9da093fe7: Pull complete
Digest: sha256:81a26c1cbe2bbd913ea1907366fb92d9ffe31eca3d2d37f144837acb0ca68ae8
Status: Downloaded newer image for nvidia/cuda:9.0-cudnn7-devel-ubuntu16.04
---> 7263fb84069a
Step 2/14 : RUN echo "deb http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64 /" > /etc/apt/sources.list.d/nvidia-ml.list
---> Running in 93c593a359bb
Removing intermediate container 93c593a359bb
---> 48c158f834bd
[...]
warning: no previously-included files matching '*.py[co]' found under directory '*'
writing manifest file 'torchvision.egg-info/SOURCES.txt'
Copying torchvision.egg-info to /opt/conda/envs/pytorch-py3.6/lib/python3.6/site-packages/torchvision-0.2.0-py3.6.egg-info
running install_scripts
writing list of installed files to '/tmp/pip-wapibxoy-record/install-record.txt'
Running setup.py install for torchvision: finished with status 'done'
Removing source in /tmp/pip-1cuu5oqj-build
Successfully installed pillow-5.0.0 torchvision-0.2.0
Cleaning up...
Removing intermediate container ceb6acafa40f
---> 45a996e31703
Step 13/14 : WORKDIR /workspace
Removing intermediate container 1e2b48d4fff5
---> b5c02b17bbe8
Step 14/14 : RUN chmod -R a+w /workspace
---> Running in d0dad20f6815
Removing intermediate container d0dad20f6815
---> 7078c5d41329
Successfully built 7078c5d41329
Successfully tagged pytorch_cuda9:latest
# python --version
Python 3.6.4 :: Anaconda, Inc.
# python
Python 3.6.4 |Anaconda, Inc.| (default, Dec 21 2017, 21:42:08)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/workspace/torch/__init__.py", line 14, in <module>
from .version import __version__
ModuleNotFoundError: No module named 'torch.version'
>>> print(torch.__version__)
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
NameError: name 'torch' is not defined
>>> import torch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "/workspace/torch/__init__.py", line 14, in <module>
from .version import __version__
ModuleNotFoundError: No module named 'torch.version'
>>>
Apparently this is an issue with nvidia-docker, as a vanilla docker launch doesn’t squawk:
$ docker run -it pytorch_cuda9:latest python
Python 3.6.4 |Anaconda, Inc.| (default, Dec 21 2017, 21:42:08)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import torch
>>> print(torch.__version__)
0.4.0a0+05908e8
>>>Looks like you are in the pytorch source directory, change directory and you should be ok.
I changed the directory but I am still getting an import error. Maybe sys.path is incomplete:
pytorch_cuda9:latest:
# python
Python 3.6.4 |Anaconda, Inc.| (default, Dec 21 2017, 21:42:08)
[GCC 7.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import pytorch
Traceback (most recent call last):
File "<stdin>", line 1, in <module>
ModuleNotFoundError: No module named 'pytorch'
>>> import sys
>>> print (sys.path)
['', '/opt/conda/envs/pytorch-py3.6/lib/python36.zip', '/opt/conda/envs/pytorch-py3.6/lib/python3.6', '/opt/conda/envs/pytorch-py3.6/lib/python3.6/lib-dynload', '/opt/conda/envs/pytorch-py3.6/lib/python3.6/site-packages']
pytorch/pytorch:latest:
# python
imporPython 3.5.2 |Continuum Analytics, Inc.| (default, Jul 2 2016, 17:53:06)
[GCC 4.4.7 20120313 (Red Hat 4.4.7-1)] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> import sys
>>> print (sys.path)
['', '/opt/conda/envs/pytorch-py35/lib/python35.zip', '/opt/conda/envs/pytorch-py35/lib/python3.5', '/opt/conda/envs/pytorch-py35/lib/python3.5/plat-linux', '/opt/conda/envs/pytorch-py35/lib/python3.5/lib-dynload', '/opt/conda/envs/pytorch-py35/lib/python3.5/site-packages', '/opt/conda/envs/pytorch-py35/lib/python3.5/site-packages/setuptools-27.2.0-py3.5.egg', '/opt/conda/envs/pytorch-py35/lib/python3.5/site-packages/torchvision-0.1.9-py3.5.egg']
Why are you importing pytorch? shouldn’t you be importing torch in the python interpreter?