add Llama 3.2 support

[tianyu-l]

Tasks

Beta Give feedback

1. Add a ViT Encoder to TorchTitan #589
   CLA Signed +1
2. [Multimodal model] Add the decoder of the multimodal model #626
   CLA Signed +1
3. Numerically stable validation for ViT
4. Numerically stable validation for decoder
5. Enable dataloading for dataset OBELICS and update the tokenizer
6. enabling loading the pretrained checkpoints for ViT and non-crossed layer in decoder
7. Training receipe for MM model
8. Perf and numerically stability for 1D (FSDP) for MM model

[TJ-Solergibert]

Hi @tianyu-l, is anyone working on the Enable dataloading for dataset OBELICS and update the tokenizer task? Let me know if I could contribute to this task!

[tianyu-l]

Hi @TJ-Solergibert, it'd be great if you can help on that. I can help discuss and review the code / design / etc. You may create a separate issue (and add the link above) to track progress if needed.

[fduwjj]

@TJ-Solergibert you can add me as the code reviewer as well.