ComputerWit Profile
ComputerWit

@ComputerWit

Followers
28
Following
4
Media
0
Statuses
5

Machine Learning Engineer at @tryolabs. I also like logic, compilers, programming languages, and the occasional philosophy of language.

Uruguay
Joined December 2020
Don't wanna be here? Send us removal request.
@ComputerWit
ComputerWit
4 years
Say you have a classification model with an accuracy of 98%, while the current manual system you're replacing is 98.7% accurate. Is the model good enough to put in production? Obviously that's not enough information, but what *is* enough information?
compwit.wordpress.com
You are developing a Machine Learning model for an application in the industry. How do you measure the model’s performance? How do you know whether the model is doing its job correctly? In my…
0
0
0
@guilleripa
guille
5 years
After working with @PyTorch's TorchServe for a bit (And doing a full 360 love it, hate it, love it cycle) I wrote on how to customize it so people could extend it even further. And hopefully spare a few headaches. Go check it out! https://t.co/bm4LfdRtYQ
Tweet card summary image
pento.ai
Learn how to develop advanced custom handlers for PyTorch's TorchServe. This guide walks you through creating tailored inference handlers, managing model artifacts with torch-model-archiver, and...
0
2
5
@emasquil
elias
5 years
Yesterday I found myself googling how to download multiple files from #AWS S3 in parallel using #Python After not finding anything reliable and spending a couple of hours on it, I wrote a post about what I learned and how to do it https://t.co/ZmdaMwfbAz
emasquil.github.io
Yesterday I found myself googling how to do something that I’d think it was pretty standard: How to download multiple files from AWS S3 in parallel using Python? After not finding anything reliable...
2
6
24
@ComputerWit
ComputerWit
5 years
Searching for optimal classification thresholds (e.g. for building decision trees) can be done neatly in log-linear time. The underlying idea of how to do it turns out to be useful in a bunch of unexpected cases.
compwit.wordpress.com
Binary classification problems (target/non-target) are often modeled as a pair $latex (f, \theta) &s=1$ where $latex f : \mathbb{R}^D \to [0, 1] &s=1$ is our model, which maps input vectors…
0
1
1