Zach

Try removing the gradio crap and making it print raw first. I'm assuming this is being caused by gradio plugin

Have you tried the following? pip install tensorflow-io

Its LlamaForCausalLM and LlamaTokenizer

Since this has been re-opened are we going to see Meta re-evaluate the license for this model?

Can anyone help with this? I need this to work as well!

Currently using torch: 2.4.1+cu24 liger-kernel: 0.3.1 triton: 3.0.0 Am using this in a DDP setup and using torch.compile. The code works and model works (training + inference) however the torch...

Would you happen to have an ETA by any chance? I do not mind installing the beta versions and trying it out when you have them

Awesome, yes it's a weird scenario because if you add any liger components and also do torch.compile then it does train but it goes to all hell during validation. My...