It usually depends on the model. When i tried nous-hermes model, it used approximate ~5GB extra RAM. I use a system with 16GB ram.
A privacy and FOSS enthusiast. Linux 🐧 and Firefox 🦊 for life
🔭 Loves space and starwatching
🌌 Amazed at how universe works
It usually depends on the model. When i tried nous-hermes model, it used approximate ~5GB extra RAM. I use a system with 16GB ram.
This solves one of the most requested ideas in mozilla connect. Yaay
I can see them in inbox, and in case of notifications, they only appear when the client is in foreground.
I can see replies to my comments and posts in infinity. The main issue is now that private messages aren’t implemented. Depending on the speed of developement, it will soon land
On android, i wish Xplore file manager had an alternative. Didn’t find anything to replace it. Dualpane file managers are the best. (btw, its one of the app where donating to dev really feels like he deserves it ).
If you want a completely offline (local) one, you can take a look at gpt4all
For PC’s the whisper can be now run on many common laptops and desktops ( with almost good accuracy ) , thanks to projects like whisper.cpp and faster-whisper.
There also happens a lot of model’s getting released by meta like wav2vec2 and MMS ( not seen using anywhere though )
I guess there are many python packages for aforementioned projects too.