Ipython compatibility with pytorch 0.4.1


I would like to ask if pytorch 0.4.1 with cuda 9.2 work with Ipython 7.0+ ? I use conda install. I was asked to downgrade my pytorch and torchvision when I tried to conda install ipython.

$ conda install ipython
Solving environment: done

Package Plan

environment location: /home/user/.conda/envs/pytorch

added / updated specs:
- ipython

The following packages will be downloaded:

package                    |            build
pytorch-0.1.12             |py36cuda7.5cudnn6.0_1       135.1 MB
mkl-2017.0.4               |       h4c4d0af_0       171.1 MB
prompt_toolkit-2.0.7       |           py36_0         482 KB
ipython-7.1.1              |   py36h39e3cac_0         1.1 MB
libtorch-gpu-0.1.12        |                0       118.9 MB
                                       Total:       426.6 MB

The following NEW packages will be INSTALLED:

backcall:         0.1.0-py36_0                               
decorator:        4.3.0-py36_0                               
ipython:          7.1.1-py36h39e3cac_0                       
ipython_genutils: 0.2.0-py36_0                               
jedi:             0.13.1-py36_0                              
parso:            0.3.1-py36_0                               
pexpect:          4.6.0-py36_0                               
pickleshare:      0.7.5-py36_0                               
prompt_toolkit:   2.0.7-py36_0                               
ptyprocess:       0.6.0-py36_0                               
pygments:         2.2.0-py36_0                               
traitlets:        4.3.2-py36_0                               
wcwidth:          0.1.7-py36_0                               

The following packages will be UPDATED:

libtorch-gpu:     0.1.12-nomkl_0                              --> 0.1.12-0                    

The following packages will be DOWNGRADED:

blas:             1.0-openblas                                --> 1.0-mkl                     
mkl:              2019.0-118                                  --> 2017.0.4-h4c4d0af_0         
pytorch:          0.4.1-py36_cuda9.2.148_cudnn7.1.4_1 pytorch [cuda92] --> 0.1.12-py36cuda7.5cudnn6.0_1
torchvision:      0.2.1-py36_1                        pytorch --> 0.1.8-py36_0

I am not using conda on that machine, but I have IPython 7.1.1 installed alongside with PyTorch 0.4.1, and I can import PyTorch and use it just fine in an IPython session. I don’t think there should be a problem.

1 Like

I installed ipython using pip from conda.
I’m using pytorch 0.4.1 from conda as well.

➜  goelakash conda list torch --verbose
# packages in environment at /Users/goelakas/miniconda3:
# Name                    Version                   Build  Channel
pytorch                   0.4.1           py36_cuda0.0_cudnn0.0_1    pytorch
torch                     0.4.1                     <pip>
torchvision               0.2.1                    py36_1    pytorch
➜  goelakash conda list ipython --verbose
# packages in environment at /Users/goelakas/miniconda3:
# Name                    Version                   Build  Channel
ipython                   6.5.0                     <pip>
ipython-genutils          0.2.0                     <pip>
ipython_genutils          0.2.0            py36h241746c_0
1 Like


I managed to solve the issue by install ipython in pip.