asesidaa

You can try bump the version of libgdx (and other dependencies), and see if that fix the issue.

Now that LWJGL3 has become the default for libgdx and LWJGL2 has become legacy, I think it's time to migrate to LWJGL3. I have experimented a little bit, and there...

> Hello @aesidaa, thanks for opening this issue. > > Just a quick note to either my future self, or anyone taking up implementation: > > * MIME type providers...

It seems that the issue can be solved by bumping ASM8 to ASM9, and the resulting build can pass the tests, but I am not sure if there are other...

I am trying to look into this, can you give some advice on working out what fields in the response need what kind of data? Are there any kind of...

I think the easiest way is to build without the `-g` flag, otherwise, you need to add `DILabel` support manually in llvm-hs as specified in the issue that you mentioned....

Can confirm this is a regression that is introduced in v0.1.33, rollback to v0.1.32 can load large models

This may also be caused by cuda version If I use a locally compiled version (with cuda 12) it can load large models just fine

Have you tried to compile ollama locally with native cuda libraries This does fix this issue on my end