9/29/23: BL Proposal Defense

 

Machine Learning Study Group

Welcome! We start at 4:00, and go for 40 minutes. New members are always welcome and anyone can join. You may attend any or all sessions. Feel free to ask to be removed from the invite list at any time, as we have no wish to send unneeded emails which we all certainly get too many of.

 

Agenda & Minutes

 

Welcome to the 82nd meeting, Sept. 29, 2023

•  Today: BL proposal defense at a different videoconference link. See the calendar entry for the link.
  
• Next week: We will be in "week" 2 of the 3-week quasi-cycle (which consists of reading in week 1, and HuggingFace labs in weeks 2 and 3, interspersed with weeks with other activities, like next week's proposal). This means we will do a Huggingface lab. Then we will do more on it in week 3, next time.
• The week after that - week 3 of our quasi-cycle with more on the HuggingFace lab we've been working through.
  

• Status of the threads of activity:
• Here is where we are with the Huggingface lab exercises:
• To continue with the Huggingface lab program, last time we got up to the section "Mask Filling" in chapter 1,  part 3, https://huggingface.co/learn/nlp-course/chapter1/3?fw=tf. We can start with that next time we do it, reading about it briefly and then trying it in the running code window.
  

• Readings: We will discuss the top ranked item from the page of evaluated items next time we do readings.

 

9/22/23: Guest Presentation on Multithreading - Thanks, EG!

 

Machine Learning Study Group

Welcome! We start at 4:00, and go for 40 minutes. New members are always welcome and anyone can join. You may attend any or all sessions. Feel free to ask to be removed from the invite list at any time, as we have no wish to send unneeded emails which we all certainly get too many of.

 

Agenda & Minutes

• Welcome to the 81st meeting, Sept. 22, 2023.
• Status updates/news/inputs/comments
• Student and project updates? 
  
• Schedule of activities:
• Today: Guest presentation by EG on multithreading.
  
• Next week: BL proposal defense at a different videoconference link.
• Following week: We will be in "week" 2 of the 3-week quasi-cycle (which consists of reading in week 1, and HuggingFace labs in weeks 2 and 3, interspersed with weeks with other activities, like next week's proposal). This means we will do a Huggingface lab. Then we will do more on it in week 3, next time.
• The week after that - week 3 of our quasi-cycle with more on the HuggingFace lab we've been working through.
  

• Status of the threads of activity:
• Here is where we are with the Huggingface lab exercises:
• To continue with the Huggingface lab program, last time we got up to the section "Mask Filling" in chapter 1,  part 3, https://huggingface.co/learn/nlp-course/chapter1/3?fw=tf. We can start with that next time we do it, reading about it briefly and then trying it in the running code window.
  

• Readings: We will discuss the top ranked item from the page of evaluated items next time we do readings.

 

9/15/23: HuggingFace lab

 

Machine Learning Study Group

Welcome! We start at 4:00, and go for 40 minutes. New members are always welcome and anyone can join. You may attend any or all sessions. Feel free to ask to be removed from the invite list at any time, as we have no wish to send unneeded emails which we all certainly get too many of.

 

Agenda & Minutes

• Welcome to the 80th meeting, Sept. 15, 2023.
• Status updates/news/inputs/comments
• Student and project updates? 
• BL will defend his PhD proposal in two weeks, Friday 9/29, at 4:00. The videoconference link is different. It will be at: 
• Google Meet, https://meet.google.com/dib-fyhb-odw.
• Therefore, we will meet next  week as usual, but the following week we will not meet at the usual Zoom link like we are today, but instead everyone is invited to the Google Meet link for BL's proposal defense.
  
• Schedule of activities:
• We are now in "week" 2 of the 3-week quasi-cycle (which consists of reading in week 1, and HuggingFace labs in weeks 2 and 3, interspersed with weeks with other activities, like next week's proposal). This means we will do a Huggingface lab today. Then we will do more on it in week 3, next time.
• Next week we will take a break from the quasi-cycle with a presentation by EG on multithreading.
• The following week will be the ML proposal (at a different videoconference link).
• The week after that - back to week 3 of our quasi-cycle with more on the HuggingFace lab we've been working through.
  

• Status of the threads of activity:
• Here is where we are with the Huggingface lab exercises:
• To continue with the Huggingface lab program, this time we got up to the section "Mask Filling" in chapter 1,  part 3, https://huggingface.co/learn/nlp-course/chapter1/3?fw=tf. We can start with that next time we do it, reading about it briefly and then trying it in the running code window.
  

• Readings: We will discuss the top ranked item from the page of evaluated items next time we do readings.

 

9/8/23: Finish evaluating a list of candidate readings

 

Machine Learning Study Group

Welcome! We start at 4:00, and go for 40 minutes. New members are always welcome and anyone can join. You may attend any or all sessions. Feel free to ask to be removed from the invite list at any time, as we have no wish to send unneeded emails which we all certainly get too many of.

 

Agenda & Minutes

• Welcome to the 79th meeting, Sept. 8, 2023.
• We are continuing to finish up, from last time, "week" 1 of the 3-week quasi-cycle (which consists of reading in week 1, and HuggingFace labs in weeks 2 and 3, interspersed with weeks with other activities). This means we will read more abstracts and evaluate whether to read the fuller articles.
  
•  Status updates/news/inputs/comments
• Student and project updates? 
• 
  
• Schedule of activities:
• Today: read & evaluate various possibilities for the next reading. We got up t0 #14 last time and will continue with #14 this time.
• Any other updates/news/inputs or comments?

• Status of the threads of activity:
• Here is where we are with the Huggingface lab exercises:
• To continue with the Huggingface lab program, last time we did this we worked on chapter 1,  part 3, https://huggingface.co/learn/nlp-course/chapter1/3?fw=tf. After the first paragraph we went to the collab link where we worked on that, starting to experiment with the code window

from transformers import pipeline

 

generator = pipeline("text-generation", model="distilgpt2")

generator(

    "In this course, we will teach you how to",

    max_length=30,

    num_return_sequences=2,

)

Next time we can experiment a bit more and then go on from there.

• Readings: let us discuss and evaluate the abstracts of the potential readings in the page of reading possibilities.
• We have evaluated items 1-11 and will finish the list next week (instead of going on to week 2 of the 3-week quasi-cycle).
  

• Readings/videos/demos we have finished. 
• 6/30/23: We finished https://openai.com/research/gpt-4 about GPT-4!
• 2/17/23; "Deep Learning for AI" by Bengio, LeCun and Hinton, the Turing Award Lecture paper for 2018, published in CACM, 2021, https://dl.acm.org/doi/pdf/10.1145/3448250.
• ChatGPT Is a Tipping Point for AI (Harvard Business Review). Finished 1/27/23.
• 12/16/22: We demoed ChatGPT.
• 11/4/22: We viewed and discussed a transformers video: https://www.youtube.com/watch?v=TQQlZhbC5ps first. This seemed to be one of the better videos of the many out there.
• Transformers From Scratch," https://e2eml.school/transformers.html. Finished on 10/28/22.
• The Narrated Transformer Language Model, Jay Alammar, https://www.youtube.com/watch?v=-QH8fRhqFHM, finished 7/22/22.
• https://www.youtube.com/watch?v=F1X4fHzF4mQ. Finished 8/26/22.
• We also finished other things from before this list of finished sources was created to keep a record of them.

 

9/1/23: Discussion and reading

 

Machine Learning Study Group

Welcome! We start at 4:00, and go for 40 minutes. New members are always welcome and anyone can join. You may attend any or all sessions. Feel free to ask to be removed from the invite list at any time, as we have no wish to send unneeded emails which we all certainly get too many of.

 

Agenda & Minutes

• Welcome to the 78th meeting, Sept. 1, 2023.
• We are finishing up, from last time, "week" 1 of the 3-week quasi-cycle (which consists of reading in week 1, and HuggingFace labs in weeks 2 and 3, interspersed with weeks with other activities). This means we will read more abstracts and evaluate whether to read the fuller articles.
  
•  Status updates/news/inputs/comments
• Student and project updates? 
• Congratulations to new Dr. VW!
• Discuss convergence issue in a NN.
  
• Schedule of activities:
• Today: read & evaluate various possibilities for the next reading. We got up t0 #14 and will continue next time.

  

The meeting ended here

• Any other updates/news/inputs or comments?

• Status of the threads of activity:
• Here is where we are with the Huggingface lab exercises:
• To continue with the Huggingface lab program, last time we did this we worked on chapter 1,  part 3, https://huggingface.co/learn/nlp-course/chapter1/3?fw=tf. After the first paragraph we went to the collab link where we worked on that, starting to experiment with the code window

from transformers import pipeline

 

generator = pipeline("text-generation", model="distilgpt2")

generator(

    "In this course, we will teach you how to",

    max_length=30,

    num_return_sequences=2,

)

Next time we can experiment a bit more and then go on from there.

• Readings: let us discuss and evaluate the abstracts of the potential readings in the page of reading possibilities.
• We have evaluated items 1-11 and will finish the list next week (instead of going on to week 2 of the 3-week quasi-cycle).
  

• Readings/videos/demos we have finished. 
• 6/30/23: We finished https://openai.com/research/gpt-4 about GPT-4!
• 2/17/23; "Deep Learning for AI" by Bengio, LeCun and Hinton, the Turing Award Lecture paper for 2018, published in CACM, 2021, https://dl.acm.org/doi/pdf/10.1145/3448250.
• ChatGPT Is a Tipping Point for AI (Harvard Business Review). Finished 1/27/23.
• 12/16/22: We demoed ChatGPT.
• 11/4/22: We viewed and discussed a transformers video: https://www.youtube.com/watch?v=TQQlZhbC5ps first. This seemed to be one of the better videos of the many out there.
• Transformers From Scratch," https://e2eml.school/transformers.html. Finished on 10/28/22.
• The Narrated Transformer Language Model, Jay Alammar, https://www.youtube.com/watch?v=-QH8fRhqFHM, finished 7/22/22.
• https://www.youtube.com/watch?v=F1X4fHzF4mQ. Finished 8/26/22.
• We also finished other things from before this list of finished sources was created to keep a record of them.