Johannes Gäßler

Results 44 issues of Johannes Gäßler

As far as I can tell there is no command that lets you move a database to a new directory after you have already created it. I think this would...

When you type `tmsu-fs-mv src1 src2 dest` like you would with the regular mv command, ~~only one of the source files gets moved to the destination.~~ it does not work....

Currently Radon is obtained from Plutonium-239. While the decay chain of Plutonium does contain Radon, it's presumably the wrong Isotope. Radon-223 has an average lifespan of 5.7s, which would be...

enhancement
muh realism

Consider the following screenshot: ![Screenshot_20230208_164020](https://user-images.githubusercontent.com/18492268/217578064-a35dace6-5a7b-467c-b57f-48ee2e5b7286.png) Presumably the leftmost button should be saying "Not English" rather than "Not Russian". I'm assuming this is a simple oversight where instead of the target...

website

Consider the following screenshot: ![Screenshot_20230212_143428](https://user-images.githubusercontent.com/18492268/218314504-0075a677-e3e4-46f7-a1fb-8555fd005fec.png) It seems that if a single word is too long for a line break the containers for individual messages increase their bounds beyond what is...

The PR https://github.com/LAION-AI/Open-Assistant/pull/1362 updated the guidelines. However, the ? icon when providing data still points at https://projects.laion.ai/Open-Assistant/docs/guides/prompting when it should now be pointing at https://projects.laion.ai/Open-Assistant/docs/guides/guidelines . (Initially I accidentally created...

Consider the following screenshot: ![Screenshot_20230217_183718](https://user-images.githubusercontent.com/18492268/219729062-eca362fc-8ed3-46f9-9542-b1f646462e10.png) Presumably the message with the code blocks should be no wider than the other ones but it extends onto the other messages on the right.

This PR supersedes https://github.com/ggerganov/llama.cpp/pull/1607 ; that one has issues with a decline in prompt processing speed. This PR adds more things on top and should end up being at least...

This PR is quite large. Its primary goal is to lay the groundwork for the implementation of further CUDA kernels for ggml operations. I am also adding multi GPU support...

enhancement
performance
high priority
hardware
refactoring

With my GPU changes in https://github.com/ggerganov/llama.cpp/pull/1703 Windows users were reporting to get garbage outputs: https://github.com/ggerganov/llama.cpp/issues/1735 . I think the problem is a bug in the Windows CUDA compiler regarding instruction...