add Llama 3.2 support

[tianyu-l]

### Tasks
- [ ] https://github.com/pytorch/torchtitan/pull/589
- [ ] https://github.com/pytorch/torchtitan/pull/626
- [ ] Numerically stable validation for ViT
- [ ] Numerically stable validation for decoder
- [ ] Enable dataloading for dataset OBELICS and update the tokenizer
- [ ] enabling loading the pretrained checkpoints for ViT and non-crossed layer in decoder
- [ ] Training receipe for MM model
- [ ] 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.

[DhruvaBansal00]

Hi @tianyu-l and @fduwjj, I would love to help out here as well. I would love to learn more about the status of the tasks listed above and which one would be the best to get started with.