As someone that is familiar with using nvidia-smi to track util, what are some commands people use to track the SM efficiency? The end of the article had some references, but no examples of what to use explicitly.
At what point is google liable for the increased rate of accidents from features like this? Surely they have some data that can estimate users getting in accidents while using gmaps.
Hey Jeremy, it seems like you could calculate exactly how much a model learns in a single step by calculating the loss for a batch a second time (with no_grad) after the loss is calculated the first time and gradients are updated. This seems like it could produce interesting outputs when graphing the difference of first and second losses at the batch or observation/question level.
I wanted to do this but never figured out where I could get access to the data. I think with setting up the inputs correctly to handle 0-5 heroes chosen per team it could work. Once you have a model you just need to rank remaining heroes by the expected win probability if they are included with the team. If you have a way to get data I would be interested.
The formula still works for scales of 5 or 10, you just have to divide by the max rating first and then multiply by it again at the end.
For example a 3/5 stars turns into 0.6 positive and 0.4 negative observation. Following the formula from there will give a lower bound estimation between 0 and 1, so then you just multiple by 5 again to get it between 0 and 5.
As a data scientist using snowflake and in the market for a feature store, the snowflake streaming is only for data ingestion, not serving. It doesn't solve the problem of serving data for a low latency app.
I've was using my 13" MBP keyboard for awhile, but ended up getting a little bit of wrist cramps from using it plus the trackpad. I got this keyboard ~2 years ago after seeing a link here and I have had 0 regrets since purchasing it.
The wrist cramps went away almost immediately (which I would assume would have worked for any split keyboard with a better wrist resting angle). The learning curve took a little as the UHK has a Mod key on one side where the space bar is supposed to be and I would also sometimes try to hit keys with one hand that are on the other side of the keyboard. The UHK agent app makes it super easy to map any key from any layer (fn or mod) to something else or even a combination of keys. I pair this with Better Touch Tool app which helps me map keyboard combinations to other things like switching to apps. The last part is using the mouse with the "mouse button" in place of capslock key, which I feel has cut the need to take my hands off the keyboard by ~70%.
It was a little pricey purchase, but the quality is phenomenal and 100% worth the money. If mine broke for whatever reason today, I would buy another in a heartbeat.
Tangentially related, but google nixed android chrome's "home"/"duet" feature which allowed having the url bar at the bottom of the screen and made navigating on a phone/tablet _much_ easier. I'm somewhat baffled how they can work on a tablet and be so unaware of what a good UI looks like.
You may not have tab access for windows/dialog boxes. I think it is deactivated by default. You can activate with ctrl+f7 or system preferences > keyboard > shortcuts > click radial at the bottom for all controls.
The reason is to get a job at Amazon where you will earn $. It is part of their offer and lots of employees take it because the salary is worth the risk of impact down the road.
Yep, kill any content an ML model has predicted of high risk or unprofitable. Classic example of a regulation second order effects resulting in the opposite outcome of what was intended.