Stop copying guided ML projects (do this instead)

[@gptLearningHub]

More Resources! Review of math concepts: kzbin.info/www/bejne/hYqtmXWgo7GjZqs Gradient Descent: kzbin.info/www/bejne/h33Pga2Eq9yMia8 Linear Regression: kzbin.info/www/bejne/aKeoZHenjMl4jtE PyTorch and Neural Networks: kzbin.info/www/bejne/iabUgY2crMd0rKM First-Principles Framework (Learn Fundamentals): www.gptlearninghub.ai/first-principles-framework Beginner's Blueprint (Build Projects): www.gptlearninghub.ai/beginner-blueprint

[@tankapadia]

sorry im a pytorch peter type of guy

[@gptLearningHub]

Haha same

[@basil9633]

This is why I rebuild projects with concepts Ive already somewhat mastered

[@gptLearningHub]

You're on the right track 🤜

[@dappavidz]

1. Gradient descent and pytorch 2. Translate ml papers to code 3. Jump into two but do two points prior, a) fine tune llama b) hack together basic rag workflow

[@gptLearningHub]

Best of luck mate!

[@เจษฎาอิ่มจิตร]

Bro, Daniel's PyTorch tutorial is one of the best on KZbin. You guided everyone to learn PyTorch, but you never actually watched the tutorial yourself.

[@gptLearningHub]

Daniel's "PyTorch in a day" video is fantastic! Wasn't bashing on any specific creators, just emphasizing the need to learn the fundamentals as fast as possible, so that one can start getting their hands dirty with papers sooner as well.

[@OPGAMING-rq5up]

Should I not learn linear algebra stats calculus probability first?

[@adityapatil318]

you should

[@zeusolympus1664]

Just the basics should do it. You dont have to cover advanced concepts right off the bat.

[@gptLearningHub]

Linear algebra, calculus, and prob/stat are definitely essential for ML. That being said, you don't have to master these before getting started with the actual ML concepts! Just the basics like matrix multiplication, basic derivatives, should be enough to get started.

[@gptLearningHub]

@@zeusolympus1664 Couldn't have said it better myself!

[@aafimalek]

where do I learn pytorch ?

[@nemesis2477]

I think the finetuning part has been replaced by automation. Om the google playground you can literally fine tune any google llm model with your data with just few clicks

[@gptLearningHub]

Yup, there's no-code tools for fine tuning now. I still advocate for fine tuning projects though, since understanding each line of code teaches a ton!

[@adityashukla9840]

I Know tensorflow I Created many projects Related to it should i switch to pytorch

[@SusBinggusNigusImpostorus]

this is not kpop video or cocomelno.. don't use 2~3sec cut edit

[@picklenickil]

Time for self reflection maybe?

[@gptLearningHub]

What sort of reflection are you thinking?

[@picklenickil]

@@gptLearningHub the kind that involves self. Maybe go back to basics.. build Shallow network, layer by layer..? A simple RL network maybe, Or take a moonshot and try solving ARC or something.. idk mate.. I used to like your videos, now they are just the same as others in my feed

[@gptLearningHub]

@@picklenickil Building a shallow network layer by layer, with just NumPy, was highly instructive for myself when getting started. I appreciate the honest feedback mate.

[@picklenickil]

@@gptLearningHub you are very welcome. Try your luck on the ARC prize. It's fun to say the least and most probably you'll fail like the reso is but it's damn fun.

[@Chadpritai]

I would recommend to implement autograd in python ❤

[@elievanazokom4809]

all your video = pytorch

[@gptLearningHub]

After learning topics like Gradient Descent, Linear Regression, FeedForward Neural Networks, I think learning PyTorch is incredibly beneficial!

[@elievanazokom4809]

you confuse ML and Deep Learning

[@gptLearningHub]

Classical ML is definitely still critical in 2024 - However I believe beginners can get started with Deep Learning before learning Classical ML, as long as they review the required maths, and master topics like Gradient Descent, Linear Regression, etc and can come back to Classical ML topics like SVM, K Means, etc later on

[@ten_cents]

all the same concept. machines be learnin