i’m very skeptical that that’s a better decision than just using onnx wasm or whatever else is available
@jxmnop
-
Reinventing Transformers Architecture From First Principles
By
–
yep hahaha he re-invented Transformers from first principles I suppose
-
Closing the Performance Gap in AI Systems
By
–
yeah true I kind of am lmao but there’s a Twitter thread where ppl are trying to close the performance gap
-
The Satisfaction of Iterative Code Improvement and Optimization
By
–
there’s a very specific satisfaction in working diligently on the same code for a long while; every day I fix something small that irritates me, or add a metric that is useful every once in awhile, or make a certain piece just a bit faster over time these micro-optimizations
-
llm.c versus PyTorch: speed and simplicity comparison
By
–
…"in that llm.c will never be as fast nor as simple as pytorch"
-

Low-Level Code Performance Optimization Zeitgeist
By
–
well that’s exactly what I said; anyway i’m referring to this tweet about performance optimizations, as well as the greater zeitgeist where everyone things it’s faster to write low-level code
-
LLM.c Versus PyTorch: Practical Limitations of Low-Level Implementation
By
–
my take on all this llm.c stuff is that it’s very impressive, and karpathy is certainly brilliant, but it was sort of a futile exercise in that llm.c will never be as fast nor as simple as pytorch this whole movement to “write everything in the lowest language possible” is a tad
-
AI Comedy: When Machine Humor Falls Emotionally Hollow
By
–
this was almost the first computer program to ever intentionally laugh (and i'm easy to make laugh) but sadly, no; i thought this was pretty dumb; it really gets the tone of standup right, though, the pauses and the cadence and everything, but the humor was really hollow
-
Trainer Framework Complexity: Overriding Methods and Logging Issues
By
–
on the other hand you can be someone like me who chose it for ease of use at the beginning of a project and now has overriden almost every single method of the Trainer and has to go to insane lengths to understand basic functionality like logging
-
Easy-to-use platform with hosted datasets and pretrained models
By
–
hey you said it! they make getting started really easy 🙂 and their hosted resources are so nice, like datasets & pretrained models