rnllama/rn-llama.h file not found in new 0.5.0 version

[MekkCyber]

Hi ! With the new 0.5.0 version I have the following error :

Any idea what could be the reason ? Thanks for the help

PS : seems related to this update : #112

[jhen0409]

Could you able to check if rnllama.xcframework is correctly imported in your project? Usually it will be imported by pod install.

Pods:

[MekkCyber]

Yes it was a pod install problem ! sorry i'm new to this 😅 and thank you for the help ! However I am getting gibberish from the model when i use it in my app any idea why ?

[jhen0409]

Yes it was a pod install problem ! sorry i'm new to this 😅 and thank you for the help ! However I am getting gibberish from the model when i use it in my app any idea why ?

Which device and model & quantization do you use?

[MekkCyber]

I am using an ios simulator with an fp16 gguf deepseek-r1-qwen 1.5B model

[jhen0409]

OK, I did see the issue with the model on iOS simulator (it skipped Metal), but it works on real device.

Also, the main issue seems be caused by llama.cpp, I can reproduce with -DGGML_METAL=OFF on llama-cli. I'll open another issue to track it.