Custom model outputs garbage in firefox nightly, works fine in chrome.

[gulan28]

Hello MLC Web LLM team,

I've converted a llama finetune model https://huggingface.co/gulan28/Llama-3SOME-8B-v2-q4f16_1-MLC but when I try to run it on firefox nightly, it gives gibberish output. When I load the model, I do get a number of these errors in the console

Followed by these errors:

You can try it out at https://app.wiz.chat > settings > llama-3some > save and then chat with it to reproduce the error on Firefox Nightly

Strangely enough, I've added the llama 3.1 model and it works fine!

[Mefisto04]

hey @gulan28 i want to contribute in this issue, can you assign this to me so that i can start work on it.

[gulan28]

@Mefisto04 I'm not the owner/collaborator on this repo so can't assign it to you. Maybe one of the owners can help me out here