Flash attention is not there yet mate. @grok
What do you think?
Replying to @YiMaTweets
A recorded youtube video would be awesome for those who couldn’t make it in person
Parsa retweeted
I'd like to start 2024 with some thoughts on the state of optimization theory. To begin with, let me quote Nesterov's 2003 book:
"The main fact, which should be known to any person dealing with optimization models, is that, in general, the optimization problems are unsolvable"
🧵
Parsa retweeted
Fourier transform on groups turns convolution into multiplication. For non-commutative groups, it is matrix-matrix multiplication though … en.wikipedia.org/wiki/Fourie… en.wikipedia.org/wiki/Convol…
Replying to @sirbayes @predict_addict
What are the advantages of this new book by Simon Prince over
Goodfellow’s DL book?
Parsa retweeted
Excited to be launching a series of short videos (5 min or less) to optimize the training performance of your PyTorch models!
Covering
- mixed-precision training
- multi-GPU training strategies
- Compiling PyTorch models
- and finding good batch sizes
🔗 lightning.ai/pages/courses/d…
Replying to @docmilanfar @benedictevans
@docmilanfar
Do you have any survey paper or references in mind which outline current open problems/questions/mysteries in DL which you or your colleagues think should be addressed?
As you see people from various domains approach this differently.
Replying to @kawaiitruecrime
عكس اقاي ناصري رو كه ديدم يادم اومد اقاي ميلاد ناصري ٤-٥ سال پيش از تجربه مهاجرتش به عنوان SW حرف زد و مطالب جالبي درمورد ديدگاه الگوريتمي گفت. حتي يادم هست يك مقاله جالبي هم در لينكداين نوشته بود در مورد تجربه اماده سازي مصاحبه گوگل.روحش شاد. sokanacademy.com/podcast/میلاد-ناصری
Replying to @ADutchEngineer
I started off with C++ and it gave me a good grasp of how pointers work with regards to memory or better abstraction of Object Oriented programming.I’d say which ones better suited hugely depends on future tasks though.
Parsa retweeted
Replying to @marktenenholtz
I have not ever found better material in this topic than this course youtube.com/playlist?list=PL… from @rasbt Lectures from 8 to 12. The entre course is amazing but the model evaluation lectures are incredible.
Replying to @lexfridman
I meant what you are saying is irrelevant to this tweet. Makes no sense
Parsa retweeted
6 System Design Interview Questions
1. Instagram Design - bit.ly/3BqamCL
2. YouTube Design - bit.ly/3bbNnAN
3. URL Shortner - bit.ly/3bbNpZr
4. WhatsApp - bit.ly/3SbA9Eu
5. Parking Lot - bit.ly/3SaTyFM
4. LMS - bit.ly/3Jk9emc
Parsa retweeted
5 Software Engineering Practices to Become a Better Data Scientist by Frank Andrade buff.ly/3Irnmsb








