Suraj Subramanian

@rallen10 Thanks for raising this. I agree, this example's learning rate is (way) too high and the exploding loss resolves by setting it to something like 0.001. I'll leave this...

@suiyuan2009 it's been a while since you posted this, but in the chance that you're still interested in this I can help you with raising a PR for this

We don't have any module called "KnowledgeBase" in this project. Are you sure you're running the right code? Looks like you're pip installing the wrong package. Please follow the instructions...

Thank you for pointing this out! Even though the tokenizer has multilingual vocabulary, currently Llama3 doesn't support multilingual inference. Currently the models are officially supported for inference in English, but...

Llama2 is permissible for commercial use with an important caveat: https://github.com/facebookresearch/llama/blob/main/LICENSE#L65

Waiting on the ALF team to respond before we merge

The change helps improve readability, lgtm

We have shared scripts for finetuning and inference at https://github.com/facebookresearch/llama-recipes

Please see this thread: https://github.com/meta-llama/llama3/issues/157#issuecomment-2110497041

We recently shared scripts for finetuning and inference at https://github.com/facebookresearch/llama-recipes