Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Run MNIST Colab notebook #453

Closed
marbre opened this issue Jan 10, 2020 · 4 comments
Closed

Run MNIST Colab notebook #453

marbre opened this issue Jan 10, 2020 · 4 comments
Labels
support Request support or ask a question

Comments

@marbre
Copy link
Member

marbre commented Jan 10, 2020

Should I be able to run the colab/mnist_tensorflow.ipynb notebook (non locally)? Seems that the IREE deps at the remote site are not up-to-date:

ModuleNotFoundError                       Traceback (most recent call last)

<ipython-input-4-4524f39e0a49> in <module>()
      3 import tensorflow as tf
      4 from matplotlib import pyplot as plt
----> 5 from pyiree import compiler as ireec
      6 from pyiree import rt as ireert
      7 

ModuleNotFoundError: No module named 'pyiree'
@marbre
Copy link
Member Author

marbre commented Jan 10, 2020

Or with other words, if IREE is not in Colaboratory by default (which I assume), is there an easy way to get it installed?

@ScottTodd
Copy link
Member

cc @stellaraccident

At the moment you'll need a local runtime, launched using start_colab_kernel.py with a working Bazel development environment to run any of the Colab notebooks yourself.

There are a few ways we can make that simpler in the future:

  • Publish IREE's python bindings as packages on pip, install on-demand in the notebooks using pip install
  • Host our own runtimes (not sure how feasible this would be with public notebooks)
  • Publish a docker image (see TensorFlow colab integration #15) with our dependencies and up to date IREE python bindings

I think the first option (publishing on pip) will make the most sense in the long term. While code is still changing frequently, maybe the docker image approach would be preferred (as an alternative to setting up Bazel). GPU and Vulkan driver availability and/or SwiftShader use may complicate each of these options.

@marbre
Copy link
Member Author

marbre commented Jan 10, 2020

Thanks for clarification. I think adding a one-liner to README.md might be useful. So, I think as the code changes frequently, it's fine that it cannot be executed at the remote site, but a hint that a local runtime is a prerequisite would helpful.

@ScottTodd
Copy link
Member

Sure, I'll add some notes to the docs.

@julianwa julianwa added support Request support or ask a question and removed question labels Jun 3, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
support Request support or ask a question
Projects
None yet
Development

Successfully merging a pull request may close this issue.

3 participants