Running examples on Google Colab

approx-ml / approx

Automatic quantization library

https://approx-ml.github.io/approx/

Apache License 2.0

11 stars 1 forks source link

Running examples on Google Colab #30

Open ma7dev opened 2 years ago

ma7dev commented 2 years ago

Describe the bug When installing the library on Google Colab, Colab raises a warning for installing numpy again as it already exists in Colab by default.

WARNING: The following packages were previously imported in this runtime:
  [numpy]
You must restart the runtime in order to use newly installed versions.

Solution Build the library without installing dev dependencies.

bushshrub commented 2 years ago

We can un-depend on numpy (maybe even pytorch actually since this will install a cpu version of torch). What do you think? @sudomaze

Also related to: #26

FarisHijazi commented 2 years ago

none of the approx code is working on colab

running dir(approx) gives:

['__builtins__',
 '__cached__',
 '__doc__',
 '__file__',
 '__loader__',
 '__name__',
 '__package__',
 '__path__',
 '__spec__',
 '__version__',
 'approx',
 'equals']

and approx.approx is a function

Help on function approx in module approx.equals:

approx(item1, item2, rel_tol=1e-05, abs_tol=None)

I'm unable to access any of the core api

FarisHijazi commented 2 years ago

also the colab widget takes you to a colab that has no GPU