Adrien FAURE

Hello, did you find a solution ? The implementation seems very nice, however the key attribute will leak because we never have the possibility to free it. (Unless we keep...

Idk why, idk how but works for me. This behavior is documented somewhere ? I don't really understand what is happening here.

Hello, I am also interested in having this feature. I might be able to try to do it if you think this could be a good fit for a first...

It depends on the use case I believe, in some cases you may want to include to your makespan the platform initialization. In this case, max(finish_time) looks good. Otherwise, from...

Thank you very much for your insight, as I had trouble identifying the root of the issue alone. I will try to upgrade the package myself, as I have some...

Hi, I've taken a closer look at this issue. Currently, it seems sensible to wait for Vulkan support to be implemented in Ollama. In the meantime, I've modified the `flake.nix`...

Oh, looks like now my iGPU is now supported (680M) (https://github.com/ROCm/ROCm/discussions/2932). I will try it thank you. I tried the variable `LLAMA_VULKAN`, but the cmake still doesn't build llama with...

I will try soon, and I guess if it works I can close the issue that is specific to rocm. It might be worth opening a dedicated issue for vulkan...

I believe the [patch](https://github.com/ngi-nix/ngipkgs/blob/mob/vula-test/pkgs/by-name/vula/patch) is the culprit for the error there; it misses restoring the original intend, which was to `decode` the `v` variable. Using `v.decode()` fixes the test. ```diff...

Hi @ioerror, it looks like the issue is deeper than the patch. Do you still need my pull request ?