serovar

Does not work with e10s.

1

With e10s the note appears but you can only write without saving it and click outside does nothing, while the addon works well in a non-e10s window.

[Suggestion] Flatpak release

3

It would be convenient if this app could be released as Flatpak on Flathub, so that, in addition to be installable on every Linux distro (just like AppImages), it could...

Add a settings backup & restore mechanism

2

I think it would be very convenient, especially for users who manually select resolvers and routes and that setup Simple DnsCrypt on multiple devices.

Make resolution of extracted images configurable

3

It would be really appreciated, if possible, the addition of a setting to raise the default image quality of the produced OCRed pdf, since the default quality is suboptimal in...

[Feature request] Add an option to keep the original color of links (i.e. to not change it)

2

I love this extension since it's fast and gives very good results, but I hate that it changes the link colors on every website and I would like to have...

Support minimize to tray on Linux

4

[FireTray](https://github.com/LaserTron/FireTray) stopped working since version 5.0.78 and it's an abandoned addon, so it will probably not get any updates, so for now the suboptimal solution is to use a third...

"Failed to quantize model" with every model but the 7B one

7

With the previous version I successfully downloaded the 65B model and quantized it. With the latest version I get this error with every model but the 7B one: `bash-3.2$ ./quantize...

Feature request: support for BLOOM-like models

1

Thanks to the amazing work of the dev NouamaneTazi with [blooms.cpp](https://github.com/NouamaneTazi/bloomz.cpp) (built on top of [llama.cpp](https://github.com/ggerganov/llama.cpp)), I suppose it should be trivial to add support for it in dalai, since...

Token generation is extremely slow when using 13B models on an M1 Pro with llama.cpp, but it runs at a fine speed with Dalai (which uses an older version of llama.cpp)

28

# Expected Behavior I can load a 13B model and generate text with it with decent token generation speed with a M1 Pro CPU (16 GB RAM). # Current Behavior...