Finally got a CNN model for facial emotion recognition running on my iPhone using MLX.
Hardest part wasn’t the model, but getting data from the camera into the model
facebook friend request confirmation
2009:
1 notification: "you and X are now friends"
2017:
- FB notification
- unread item on Messenger
- email
- letter in your mailbox
- SMS to your parents
- call to your senator
I was able to implement
@karpathy
's Let's build GPT video in MLX Swift in <300 LOC.
M3 MacBook Pro achieves a loss of 1.558 compared to Karpathy's loss of 1.4873 on A100 GPU.
The model can spit out Shakespeare text like:
VIRGHARD III:
Come to thee, king we would I would seee
when my son was born, I created an email address for him.
Every achievement and photo taken is sent to that email.
When he is old enough, he will have accurately targeted ads.
Sharing this idea for those who may like it to do as well.
When my son was born, i created an email address for him.
Every achievement and photo taken is sent to that email.
When he is old enough, he will get the password and will have all those memories.
Sharing this idea for those who may like it to do as well.
Last week my PR to draw a number and infer the digit was merged in mlx-swift-examples.
I know MNIST digit prediction is Neural Net 101, but still cool for me to see it merged.
Also can't believe we can now write Neural Nets in Swift!
Paytm is a:
- Bank
- Wallet
- Ecommerce site
- Deals site
- Rechard site
- Event ticketing site
- Travel site
Pretty sure they'll do a robo investing product as well
Very inspiring. Isaac Newton was working from home due to COVID-1684 for 2 months. It is during this time that he invented gravity. Without him, we’d all be floating.
Our team recently worked on a rewrite of the Tinder app’s Purchase Layer. We share our learnings here.
Key points:
- Modular
- Undefined behavior sanitizer
- Enterprise build support for displaying paywalls
- High test coverage