This is my 4th or 5th attempt at this challenge since I originally began learning to code. I am coding and livestreaming my journey as a part of the challenge so livestreaming is just as important to me as completing my daily material.
My planned learning path: fast.ai - Practical Deep Learning for Coders -> Hugging Face - NLP Course -> DeepLearning.ai - NLP Specialization
- Improve Python expertise
- Contribute to Open Source ML library / model / project
- Notion AI/ML Hub
- Improve in Machine Learning
- Learn Deep Learning
- Focus on NLP
- Implement a research paper
- Encourage others yayy!!
Today's Progress: I decided on my planned learning path: fast.ai - Practical Deep Learning for Coders -> Hugging Face - NLP Course -> DeepLearning.ai - NLP Specialization
Thoughts: I feel good about the materials I've chosen for my learning path and plan to provide some of the resources I've found for others that are interested in DL.
Link to work: ...
Link to content: Blog: ...
Today's Progress: I watched the beginning of the Lesson 1 video and set up my Kaggle notebooks.
Thoughts: Lesson 1 covered the idea of using image recognition models for other types of data, now I'm super curious about experimenting with other types of data like sound. My first impressions on the course are all positive. I appreciate that there is a focus on top-down learning and using multiple resources to learn the same information was also mentioned. In addition to the video and text materials, I will be following the corresponding book Deep Learning for Coders with Fastai and PyTorch: AI Applications Without a PhD.
Link to work: ...
Link to content: Blog: ...
Today's Progress: I finished the video lesson, set up my Python env and got a local Jupyter server running.
Thoughts: Since I finished lesson 1 yesterday, I want to begin working through the book and get started on an image classification project.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Today's Progress: I started reading Lesson 1 from Deep Learning for Coders with Fastai and PyTorch: AI Applications Without a PhD
Thoughts: I have set up my python environment and Jupyter local server so I will work through the book materials for chapter 1 including any exercises that come up. If I finish the projects without working on the image classification example, I will do that as a follow up project. I also like that I saw a long list of Quiz questions at the end of the book. That gives me more to study from.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Today's Progress: I read more of Lesson 1 about the history of Deep Learning. There are some foundational essays, papers, and books that I am taking note of to look into for later study.
Thoughts: So far I have been refreshing on the history of Deep Learning and neural networks. It is helpful to understand how the field has evovled over time and pivotal moments that account for growth and advancement.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Today's Progress: I set up Google Colabs to run my Deep Learning for Coders with Fastai and PyTorch: AI Applications Without a PhD Lesson examples and connected them to a repo on Github.
Thoughts: Today I want to make sure I set up some cloud-based notebooks for my projects as I discovered yesterday, even the simple training is a lot with me livestreaming at the same time.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Today's Progress: I read more of the Lesson 1 material from Deep Learning for Coders with Fastai and PyTorch: AI Applications Without a PhD, but could not finish training my model example (image classifier) before the end of my coding session.
Thoughts: Yesterday, I got my notebooks running in Google Colabs. I have some other platforms I want to test out, but I've used Colabs for most of my projects recently and it works well to back them up to Github. Yesterday it took an hour for the image classifier example to train and I ended up leaving it to train after stream ended so as I work through the material today I'm going to start the training early and hopefully be able to work with the model during the streaming session.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Thoughts: I am looking into model deployment with the intention of making training and engaging with my models easier alongside my studying.
Today's Progress: I finished reading more of the Lesson 1 material and got into the breakdown of the image classifier example. I decided to wait on deployment since it is literally Lesson 2 and just work through the Lesson 1 material to finish in the next few days. I also used one of my pomodoro breaks to add something new to the AI/ML Notion Hub that I'm building so I will try to make that a habit to help finish it by the end of the month.
Link to work: Kaggle notebook adapted to my own example.
Link to content: Blog: ...
Thoughts: As I finish up my chapter 1 reading, I am listening in on Senate hearings on AI Oversight. I would like to incorporate other news and industry updates as it is important to keep up with these items and talk about them within the developer community as well.
Today's Progress: I finished watching the hearing and I was able to read through the image classification example explaination and recap it aftwards just to make sure I digested while I was multi-tasking.
Link to work: Kaggle notebook adapted to my own example.
Link to content: A YouTube Playlist for AI News:
Thoughts: As I finish up my chapter 1 reading, I am listening in on House hearings on Generative AI and Copyright Law. AI News Playlist already has another video added!
Today's Progress: I finished watching the House hearings and didn't finish a lot of reading, but I learned about examples of applying image classification to other datasets.
Link to work: Kaggle notebook adapted to my own example.
Link to content: A YouTube Playlist for AI News:
Thoughts: Today I'm proud of the progress I've made so far. As I go into another 10 day round, I would like to improve my pacing a little bit on finishing the reading materials. My goal is to finish chapter 1 today.
Today's Progress: I got to see some other common DL examples including Object Detection, Sentiment Analysis, and a Recommendation System through the reading.
Link to work: Kaggle notebook adapted to my own example.
Link to content: A YouTube Playlist for AI News:
Thoughts: In addition to the Image Recognition example, I would like to adapt all of the basic DL use cases that were shown in chapter 1 of the book into projects.
"Lab"
- Object Detection
- Recommendation System
"Mini-project
- Image Classifier
"Project"
- Sentiment Analysis
Today's Progress: I finished reading chapter 1!
Link to work: Kaggle notebook adapted to my own example.
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm going to focus on deployment and looking forward to trying some newer ML tools that I've been eyeing for a minute.
Today's Progress: I was able to finish most of the lesson 2 video and got a model deployed and running through Hugging Face Spaces and Gradio.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: I feel pretty accomplished to have an understanding of model deployment with Hugging Face and Gradio and feel more empowered to work on exercises and projects now.
Today's Progress: I finished the Lesson 2 video.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: It was exciting to see that there is an API endpoint for querying the model after deploying with Hugging Face Spaces and Gradio. I'm itching to build that into a project, but taking the time to cover the book materials and dig deeper on deployment before I start anymore training.
Today's Progress: I finished most of the introductory reading.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm feeling good about my course pace, but ready to start publishing more content. Today is a day where I want to focus on a shorter learning session so I leave some fuel in the tank to get some of the work done for my blog and podcast publishings. I also have a Git & Github talk coming up in 2 weeks and I want to film/edit some supporting materials and have them ready at the talk. Content creation (teaching and documenting my journey) are just as important to me as getting through the learning so I want to make sure I prioritize it along the way. Today I'm getting into the technical section of the reading and time-boxing myself so I can work on some of my content for the week.
Today's Progress: I finished the bear classifier training example.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: Today will be another example of deploying a trained model. This time I was able to train through Google Colabs so I will be looking into how to export the model similar to what I did in Kaggle during the video lesson.
Today's Progress: I was able to get the model exported and downloaded from Colabs and close to finishing the chapter 2 reading.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm focused on wrapping up the chapter 2 reading so I can move on to a project and start Week 3 material.
Today's Progress: I finished reading chapter 2.
Link to work: My first deployed model (cat/dog image classifier).
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm researching how to curate my data for an image recognizer and wiring up my new blog. Maybe I'll make my own RSS feed, but I need to find a framework/CMS to manage it.
Today's Progress: I decided on using an RSS feed as the source for my blog content. I was able to get it setup and hosted through GH pages as well as use Zapier to integrate it with Ghost for my blogging platform.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm focused on wrapping up the chapter 2 reading so I can move on to a project and start Week 3 material.
Today's Progress: I covered the RSS 2.0 spec and best practices.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm focused on wrapping up the chapter 3 reading so I can move on to a project and start Week 3 material.
Today's Progress: I covered the intro material for the ethics chapter of the fast.ai book.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for AI News:
Thoughts: I'm focused on wrapping up the chapter 3 reading so I can move on to a project and start Week 3 material.
Today's Progress: I read a bit but I ended up watching the State of GPT (Added to the AI news section on my youtube channel --v)
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I'm focused on wrapping up the chapter 3 reading so I can move on to a project and start Week 3 material.
Today's Progress: I continued reading.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I'm focused on wrapping up the chapter 3 reading so I can move on to a project and start Week 3 material.
Today's Progress: I continued reading.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I would like to get an overview of a GPT model to help fuel future research and projects. This presentation seems like a good one to prime me on the process.
Today's Progress: I caught Training GPT from scratch with nanoGPT, but I was late so I still want to rewatch the recording.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I caught Training GPT from scratch with nanoGPT, but I was late so I still want to rewatch the recording.
Today's Progress: I reviewed Training GPT from scratch with nanoGPT.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I reviewed Training GPT from scratch with nanoGPT yesterday. Today I'm looking forward to setting up my own code base and training experiments with Weights & Biases.
Today's Progress: I finished most of "Let's Build GPT from Scratch" by the author of the nanoGPT project.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I watched a lot of the video lecture for the nanoGPT repo and although it was insightful, I didn't internalize a lot of it. Today I'm going to find a new approach to dive into the material.
Today's Progress: I did some high level research on my next 2 projects.
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I'm gathering thoughts on starting my sentiment analysis and image classification projects.
Today's Progress: I got a MongoDB instance running in the cloud and I got Compass setup on my MacBook with an empty collection for twitch chats
Link to work: A custom RSS feed for my blog.
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I'm focusing on getting data flowing to the DB.
Today's Progress: I got my ingestion service running and deployed to render, but there is data duplication in my DB that I need to troubleshoot.
Link to work: Chat Feels - Data ingestion from Twitch chat to MongoDB
Link to content: A YouTube Playlist for the State of AI:
Thoughts: I am unsure why I am getting duplicate messages from my ingestion service so I want to troubleshoot that.
Today's Progress: I worked out project goals and other technical components needed alongside building out the README.md and went with running the project locally for now to prevent data duplication. Render is struggling with the deployment and there is a reference error with the MongoDB client that may be causing the issues. This is something to follow up on next. In the meantime I need to consider my classifications and need 100 examples per category.
Link to work: Chat Feels - Data ingestion from Twitch chat to MongoDB
Link to content: A newsletter to receive resources and updates from me:
Thoughts: I am unsure why I am getting duplicate messages from my ingestion service so I want to troubleshoot that.
Today's Progress: I researched async/await keywords and came up with a list of items to potentially solve my deployment/db client connection issues.
Link to work: Chat Feels - Data ingestion from Twitch chat to MongoDB
Link to content: A newsletter to receive resources and updates from me:
Thoughts In the midst of a busy day preparing for my son's birthday away from home and ill-prepared (without even my laptop) I still want to keep my commitment as I've been consistent thus far. I will focus on reading some research tonight and skip my stream, but my streak remains.
Today's Progress: I read part of Attention is all you need, but the reality of exhaustion made it more difficult than normal to process the information so I will be continuing this read in the future as it's a crucial one in NLP.
Thoughts I missed a day yesterday amongst the busy day and a much-needed decompress. I do wish I had prepared better to make it easier to get my hour in, but I got a good reminder why planning is so important. Planning is more for the unexpected than the expected, at least for me. Every time I think I can get by without it, something unanticipated happens and a lack of planning beforehand makes it harder to adjust in real-time without making compromises that weren't really necessary. If I had planned the weekend more thoroughly I could've done everything I wanted without sacrificing my hour or my downtime. That said, especially as a new mom on my 1st birthday and getting back into pursuing serious study for the 1st time in years, I give myself grace. It's really important for me to strike a balance in 100DaysOfCode that I can carry over into my University studies for the next 6-8 years. Better to work the kinks out now because life will always happen without a doubt and I don't want to let my education suffer or vice versa. Today I am continuing my reading and I'm already digesting the information a lot better. I decided to get into highlighting and writing notes in the document as a way to summarize for myself as I go. I'm eager to read the new Falcon paper, but it's important to me to ground myself in the history and context of Transformers 1st before I jump elsewhere. I also have a bad habit of switching tasks without completing others and I'm working that out of myself along with my tendency to procrastinate. This reflection feels appropriate for blogging, and gves me some inspiration for how to get that work going regularly as well. I think I'll add it as my next post and do more reflection moments like this in the future.
Today's Progress I covered more of the paper, but I did not finish.
Thoughts I'm slowly working my way into the week, but planning on getting some more focus time done and maybe summarize AIAYN. I know a lot of summaries and videos already exist, but I feel like this is good practice for myself as I get more comfortable with digesting research. This is also a great prep to segway back into the nanoGPT training which covers all of the core model architecture.
Today's Progress: I still didng't finish the paper, but I feel good about making incremental progress on learning.
Thoughts Finishing up my reading as I work my way back into streams and working on my project this week.
Today's Progress:
Thoughts Finishing up my reading as I work my way back into streams and working on my project this week. I decifed to take a final night offline as this is my son's birthday, but look forward to returning to community for the support and accountability. Today's Progress:
Thoughts A recent conversation on Electic Vehicle development last week during our trainX.ai podcast got me back into the idea of building an ML project for an automotive application. I once studied Mechanical Engineering, had a brief stent in an engine diagnostics research lab, and worked on a connected vehicle platform. At one point I wanted to merge these interests and I'm open to exploring some problems and projects in this space.
Thoughts I'm working in keeping up some meaningful work. I've been hesitant to work on certain items if I cannot share them on stream, but its time to consider expanding my offline learning tasks. For now I'm interested in exploring some EV resources I received recently.
Thoughts I'm feeling rejuvenated and ready to start a new week, but I need to consider an update to my working schedule. My goal is to reset my work stations before the week starts and outline some YouTube videos to talk about some things I've learned so far.
Thoughts I felt like I wanted to start off Day 40 a bit more official returning to streams and needed a break to reat and reorganize. I was off longer than anticipated, but had the opportunity to get some todo items off my playe and off my mind.
Today's Progress Today I read into some AI news and hosted Episode 4 of our trainX.ai podcast.
Thoughts Today has been a planning day for me. I reassessed my goals and timeline with a day 100 target in mind. I feel equipped to power through a 60 day sprint from today until August 31. For the sake of consistency and completion of the challenge I am embracing offline learning more as it has beeb hard to find a good time for streams around baby boy's changing schedule. I feel like I can see the finish line now and feel a sense of urgency to dig back in. Today I will read about how people digest and apply research papers in order to begin to form my own process. I'm struggling to get through longer papers and get the feeling there's a better way to approach it than trying to read 40 page papers at the end of an exhausting day. 😓
Today's Progress I found a lot of useful videos with some strategies to consider and freated a new resource playlist on YouTube.
Link to Content: Tips for Reading Research Papers
Thoughts Based on some of the videos I found yesterday, I was able to start on a new notion page summarizing them as a resource. I am putting together some of tge key takeaways and publishing the new page in the AI/ML Hub.
Link to Content: Tips for Reading Research Papers
Thoughts I'm revisiting reading and considering creatong some paper summaries after getting a little guidance on tackling research papers.
Thoughts I really want to get tgroygh a backlog of articles and blogs piling up in our Discord channel. I got through a lot last week, and am picking back up today.
Thoughts I feel like it's a good time to reevaluate my goals and set up a plan leading up to the end of the challenge. I realized that I had a goal of feeling comfortable to implement a paper at the end amongst some more specific goals that I outlined. I mighy streamline a bit and maybe deprioritize some things to get the most out of my time.