Make torch environment default in jupyterlab-pytorch image

[StanHatko]

For the jupyterlab-pytorch image, the conda environment needs to be used for the torch package to be accessible. The whole point of jupyterlab-pytorch is to run PyTorch, so the conda environment should be the default in the terminal and notebooks.

For the terminal, this can be done by automatically running conda activate torch at startup. Previously I had to run bash since the default shell was previously incompatible with conda activate, but now that's no longer necessary, conda activate is now working for me in the default shell. This will make it so the torch conda environment will always be the active one in the terminal, which means Python scripts and command-line Python will always use the correct environment (the torch one). Alternatively, after any jupyterlab setup is done, is there any other way to make the torch environment the default one in conda?

For notebooks, to use torch it's necessary to select the correct environment for the kernel, as shown in the attached screenshot. Please make Python [conda env:torch] the default kernel so PyTorch will work automatically in new notebooks.

[StanHatko]

Actually I just tested again, and conda activate is now working in the default shell. I've updated the issue to reflect that.