Remote NLP Study Group meets Saturdays at 8 AM PST, starting 12/14/2019

Thank you @foobar8675, good catch. It’s fixed now!

@jcatanza i started looking a head to video 8 and am a bit torn on using the fastai v1 library since v2 is coming soon. Do you have any thoughts on that?

Notes from the Saturday 1/25/2020 meetup discussing the notebook 3-logreg-nb-imdb_jcat.ipynb

  1. In order to get the notebook to produce the table of accuracies at the end, you have to first install the tabulate package: in a shell terminal window, run the command:
    conda install tabulate

  2. I added a brief discussion of Bayes’ Theorem to the notebook.

I think you should press on using the v1 library. My feeling is that familiarizing with v1 will ultimately make it easier to learn v2.

The Fastai NLP Study Group will meet
Saturday February 01, at 8 AM PST, 11 AM EST, 5 PM CET, 9:30 PM IST

Join the Zoom Meeting when it’s time!

Topics: Fun with Bayes' Theorem; Numerical Stability; regex (regular expressions)

Suggested homework / preparation:

  1. Watch videos #6 and #7. These two videos are relatively short (about an hour total). The lesson will focus mainly on video #7: regex (regular expressions).

Video playlist is here

  1. Read and work through notebooks 3b-more-details_jcat.ipynb and 4-regex_jcat.ipynb

  2. Note: in order to access and run the _jcat.ipynb notebooks you’ll need to clone the Study Group’s github repository >

To join via Zoom phone
Dial US: +1 669 900 6833 or +1 646 876 9923
Meeting ID: 832 034 584

The current meetup schedule is here.

Sign up here to receive meetup announcements via email.

Thank you @jcatanza. I’m still a bit torn but appreciate your thoughts.

@foobar8675 Have you installed Fastai v1?

Yes, I have.

1 Like

The Fastai NLP Study Group will meet

Saturday February 08, at 8 AM PST, 11 AM EST, 5 PM CET, 9:30 PM IST

Join the Zoom Meeting when it’s time!

Topic: Introduction to Language Modeling using Deep Learning

Suggested preparation:

  1. Watch video #8 Video playlist is here

  2. Read and work through course notebook #5

In this reworked and annotated version of notebook 5-nn-imdb.ipynb, I

  • Fixed a few minor errors, enabling the notebook to run to completion
  • Implemented a workaround for a bug in fastai's text data API that seems to affect only Windows 10
  • Organized the material into coherent sections
  • Added step-by-step explanations/annotations throughout, indicating where transfer learning comes into play.
  1. Note: in order to access and run the _jcat.ipynb notebooks you’ll need to clone our Study Group’s github repository.

To join via Zoom phone
Dial US: +1 669 900 6833 or +1 646 876 9923
Meeting ID: 832 034 584

The current meetup schedule is here.

Sign up here to receive meetup announcements via email.

The Fastai NLP Study Group will meet
Saturday February 22, at 8 AM PST, 11 AM EST, 5 PM CET, 9:30 PM IST

Join the Zoom Meeting when it’s time!

Topic: ULMFit for non-English Languages

Suggested preparation:

  1. Watch video #10
    Video playlist is here
  2. Read and work through these course notebooks
    nn-imdb-more_jcat.ipynb ,
    nn-vietnamese_jcat.ipynb , and
    nn-turkish_jcat.ipynb

Note: in order to access and run the _jcat.ipynb notebooks you’ll need to clone our Study Group’s github repository.

To join via Zoom phone
Dial US: +1 669 900 6833 or +1 646 876 9923
Meeting ID: 832 034 584

The current meetup schedule is here.

Sign up here to receive meetup announcements via email.

Can you share some links for better understanding basics of pytorch…

@Shefs0709

I would start with these two resources

This would get you going. Good Luck!! Happy Learning.

Thank you sir…:blush:

Has anyone successfully run through the 7-seq2seq-translation.ipynb notebook on
Google Collab even using the Pro version.

I am unable to get the line:

with open(path/‘giga-fren.release2.fixed.en’) as f: en = f.read().split(’\n’)

to work as Collab keeps running out of memory.

never mind - reading it line by line helped. figured it out.

I’m having trouble understanding something in 2-svd-nmf-topic-modeling. It says:

But if we had one vector with the relative frequency of each vocabulary word out of the total word count, and one with the average number of words per document, then that outer product would be as close as we can get.

The phrase “average number of words per document” seems wrong to me. I interpret that as:

a = [count_across_all_docs(word) / total_word_count for word in vocab]
b = [total_word_count / len(docs) for doc in docs]

It seems to me that it should be:

a = [count_across_all_docs(word) / total_word_count for word in vocab]
b = [word_count(doc) for doc in docs]

Is my understanding correct? If so, I would change the text “average number of words per document” to just “number of words per document”.

Or am I misunderstanding something?

Zoom link seems to be invalid https://zoom.us/5167464197 . Does anybody have the right link for the meetup ??

[Wiki] Updated the wiki with the right zoom link https://zoom.us/j/5167464197 that worked for me

1 Like

The cNLP classes authorityapk

Is GPU necessary for running pytorch ???