@thor What is it? And what kind of marketing did you use?

@trinsec yeah I'll game with it. Not sure what yet though, probably CS:GO. It'll probably be mostly a machine learning computer though. A training epoch that would have taken 15 seconds is taking less than a second now!

@trinsec yeah only had onboard GPUs before. I'm extremely cheap and typically use really old hardware (I went through most of university using an arm-based chromebook with debian on it), but old hardware has really restricted me with machine learning stuff. This is the first CUDA-capable GPU I've had.

Got a new toy today, a Geforce 1050Ti. First real graphics card ever. I'm excited to see what it can do!

15.ai/
This is actually pretty fun. Portal voices are pretty solid. The timbre of Twilight Sparkle is dead on - really surprised me. Wasn't able to change denoise setting though: might have helped with some of the muddiness in a few of the voices.

Also here's the source if anyone's interested. It was made in just a few minutes and is about as bare-bones as it gets.

github.com/lukehackworth/basic

Show thread

Does anyone know how to get a complied python binary to run properly on Windows? I made a youtube-dl GUI application and compiled it using pyinstaller, but it keeps getting flagged by Windows Defender.

For machine learning analysis of sound, I've found nothing that beats a spectrogram.
While making this post, I wondered, can you make audio from a spectrogram? The answer is, apparently you can. stackoverflow.com/questions/57

@freemo software engineer? That puffs up my chest, I'm only a Sysadmin. Though I probably should be doing software stuff. Thanks!

@Pat amazing! I'd seen it years ago, but that was well before I knew much about how a neural net worked. I'm gonna have to make a crappy clone of that one day

I've only been playing with deep learning stuff for a few months now, and it's kind of mind-blowing how fundamentally simple it is to understand with fastai.
All deep learning is this: you give the software inputs and desired outputs, and the computer creates a function that converts the input as close to the desired output as possible.
Once this is understood, most of the difficulties and bottlenecks make intuitive sense.

@r2qo @freemo I'm not an experienced AI practitioner and I don't do that, but object oriented programming(what you're talking about) allows for easier abstraction of concepts and processes, making work at a large scale easier. It may not matter when you make a script or are just playing around, but when you work with other people in one codebase, it can help a lot(from what I understand. I have no experience with it yet).

@Wetrix it's the result of a machine learning app I'm working on. I trained a U-Net(a machine learning model) to tilt eyebrows up, then I ran a picture of me through it 20 times. What you see is what happened when I did that.

I'm working on creating a GAN so I can get photorealistic results, but it isn't working yet.

I'm playing with U-Nets. You get some pretty interesting results when you pass an image into a model multiple times.

I have noticed since watching apache logs on my blog that the emails of Pleroma users are visible. Is that how it should be? Seems like it's a security leak.
Are mastodon emails de facto public too?

Here is a recent project I worked on. I used fast.ai to train a machine learning model to detect audio clips that contain the sound of typing. lhackworth.com/2020/12/18/typi

Time for my ! I'm not new to this instance, but I'm new to this account. I'm a systems administrator at an MSP in Missouri. I've used various distros of Linux for over 12 years and have an interest in and .
I look forward to getting to know you all!

Qoto Mastodon

QOTO: Question Others to Teach Ourselves. A STEM-oriented instance.

An inclusive free speech instance.
All cultures and opinions welcome.
Explicit hate speech and harassment strictly forbidden.
We federate with all servers: we don't block any servers.