0

I have more than two environments in Anaconda, all three for different purposes. But, I have a few packages like PyTorch which I need to install in all the three environments. But they are too heavy and might consume up to 3 GB data, which means I will need more than 9 GB for all 3 environments.

Is there a way to share the packages which when installed in one environment can be utilized or directly be copied to other environments so everything gets done within 3 GB? Or will I have to download the whole 9 GB(3 GB thrice) for the 3 environments?

Jayampathy Wijesena
  • 1,670
  • 1
  • 18
  • 26
Hima Siva Kalyan Reddy
  • 1
  • I don't think it's copied 3 times, but the individual environments are just linking to the package cache in `/pkg` – cel May 06 '20 at 15:19
  • Does this answer your question? [Why are packages installed rather than just linked to a specific environment?](https://stackoverflow.com/questions/55566419/why-are-packages-installed-rather-than-just-linked-to-a-specific-environment) I.e., as long as you "install" the same build that is used in each env, Conda will automatically share it via hardlinking. – merv May 07 '20 at 02:20
  • Thanks a lot! But I really don't understand the downloading part, I had to download the packages repeatedly for different environments, any way of solving it? – Hima Siva Kalyan Reddy May 07 '20 at 16:08
  • @HimaSivaKalyanReddy That’s the point of the question they linked to: Install the package as you normally would in each environment, Conda will not actually download it and duplicate the files if it can avoid it. – AMC May 09 '20 at 01:51

0 Answers0