Best way to install nightly pytorch

1

Hey Pytorch user and developers

It seems that the current nightly pytorch to up-to-date. It also does not have cuda91 dependency.
When I type in:

conda install -c pytorch pytorch-nightly

It produces:

Solving package specifications: .

Package plan for installation in environment /opt/anaconda3/envs/test:

The following NEW packages will be INSTALLED:

    intel-openmp:    2018.0.3-0                                         
    libgcc-ng:       7.2.0-hdf63c60_3                                   
    libgfortran-ng:  7.2.0-hdf63c60_3                                   
    libopenblas:     0.2.20-h9ac9557_7                                  
    libstdcxx-ng:    7.2.0-hdf63c60_3                                   
    ninja:           1.8.2-h2d50403_1                        conda-forge
    numpy-base:      1.14.3-py36h0ea5e3f_1                              
    pytorch-nightly: 2018.05.07-py36_cuda8.0.61_cudnn7.1.2_1 pytorch    

The following packages will be UPDATED:

    mkl:             2017.0.3-0                                          --> 2018.0.3-1           
    numpy:           1.13.1-py36_0                                       --> 1.14.3-py36h28100ab_2
    scipy:           0.19.1-np113py36_0                                  --> 1.1.0-py36hfc37229_0

If this channel is periodically maintained? I have had a hard time building pytorch directly from github…

2

We are having some build system issues in our nightly builds following the merge with caffe2 code. After releasing 0.4.1, we will fix that. cc @smth

3 Likes