raML - Progress
80% accuracy on MNIST. Ay oh! Pretty sure if I just ran it longer, would get to 90s. Next milestone?
80% accuracy on MNIST. Ay oh! Pretty sure if I just ran it longer, would get to 90s. Next milestone?
Was watching one of George Hotz's livestreams where he was working on tinygrad library. I was always confused about what he was doing, but then I saw this statement: "tinygrad will always be below 1000 lines. If it isn't, we will revert commits until tinygrad becomes smaller." So beautiful. Everything made sense. Tinygrad was just a bigger version of karpathy's micrograd . Anyway, this is what I ended up reading only to realize that while I was explicitely defining the derivatives in raML, micrograd seems to refer to sympy for gradients. While I was aware that sympy could do it, I found it to be slow in the past, but maybe it isn't as bad as I thought. Gonna try out. (PyTorch?) Also tinygrad defines a Tensor class which generalizes numpy's ndarray, which is also interesting. Another idea: expose model weights via some API. FastAPI maybe? Another idea 2: I should DarkMode this website
Apologies to all of my non-existent readers for not putting up any updates in the last 2 weeks - Things got quite busy! First things first - work. Got to implement my first big thing (with a relatively simple mathematical idea) - inventory throttling. "What's that?" you ask... Good try! Correction - Terrible Try! You sneaky competitor! *insert some disclosure statement about sensitive information, even though it might not be that sensitive, but still, who knows, well I kind of know, but I'll pretend not to know. MOVING ON. Onboarded my first client, jeez louise. SQL, Flyway, Postresql, Scala, aaaaaaaaaaaaaah! Then renewed another client -- aaaaaaaaaaah! Gonna have to onboard another (slightly shorter) aaaaaah. Following de mahney Alright. Spent thankgiv with fam. Went for another hike with friendos. Oh! Also finally ordered a standing desk and an ergonomic chair. 10 days ago. Was supposed to be shipped 7 days ago. I just got the table .... top today (not even legs). Edit: got the legs. No sign of a chair though On a different note, talked with Romulo this morning about his cool project. TLDR building ML models to detect extreme events (human falls and others) with audio channels. Might hear about it more in the future!
Alright alright alright! Dear ladies and gents, how is it going! Man, I was told that I'm using too many exclamation marks. Too many!?!?!? That can't be right. Anyway, ED 2.0 is out-ty! If you are Mark Cuban or any other CEO of some billion dollar company, reach out yo! Let's get you into the party wagon What else what else. Got a minor surgery done, as a bonus, I'm gonna know if I have cancer lol On a darker note, my friends are fkcngi awesome! Had some cool philosophical chats with them. Oh yeah, went out for 2 hikes in the last 2 weeks. 100% per week (-> that's the type of data scientist I am)
Releasing this information puts my life at risk, but it must be done. Similar to how no one nation should have a monopoly on nuclear power, I've decided that I must release the demo of what my evil colleagues have been working on for the past decade. If you find my dead body somewhere under the Great Bridge of Texas, I'm not surprised.