Hey guys, I'm seeing the following stack trace when attempting to get gpu acceleration working on my local environment. CPU mode works perfectly, but as soon as I install the typesense gpu deps, I get the attached stack trace / crash. My gpu is on cuda version 12.7 and I've set up my dockerfile (im running typesense through docker) as you've specified here https://typesense.org/docs/guide/install-typesense.html#using-a-gpu-optional To my knowledge, my docker setup is correct (my gpu can be accessed in the container and the required gpu libraries seem available to me? (i.e. LD_LIBRARY_PATH seems correctly setup) Just wanted to run it through you all as a last ditch attempt to see if there's anything obvious i'm missing? Otherwise will just run the cpu mode (I should also say I get the same error with the models in the typesense huggingface repo - so shouldn't be my finetuned model being the issue)

Could you try with CUDA 11.8?

I did try that, with the container cuda version being set, but saw the same behaviour unfortunately - unless you mean changing the host cuda version - ill try that later today, but that requires me to rollback my pc drivers I think

Yes, you should try with 11.8 host cuda version

That did it! Thanks a ton

You are welcome!